• Akaike, H., 1974: A new look at the statistical model identification. IEEE Trans. Autom. Control, AC-19. 716723.

  • Barlow, M., , S. Nigam, , and E. H. Berbery, 2001: ENSO, Pacific decadal variability, and U.S. summertime precipitation, drought, and streamflow. J. Climate, 14 , 21052128.

    • Search Google Scholar
    • Export Citation
  • Charney, J. G., , W. J. Quirk, , S. Chow, , and J. Kornfield, 1977: A comparative study of the effects of albedo change on drought in semi-arid regions. J. Atmos. Sci., 34 , 13661385.

    • Search Google Scholar
    • Export Citation
  • Dirmeyer, P. A., 1994: Vegetation stress as a feedback mechanism in midlatitude drought. J. Climate, 7 , 14631483.

  • Gershunov, A., , and T. P. Barnett, 1998: Interdecadal modulation of ENSO teleconnections. Bull. Amer. Meteor. Soc., 79 , 27152725.

  • Gershunov, A., , T. Barnett, , and D. Cayan, 1999: North Pacific interdecadal oscillation seen as factor in ENSO-related North American climate anomalies. Eos, Trans. Amer. Geophys. Union, 80 , 25.

    • Search Google Scholar
    • Export Citation
  • Glantz, M. H., 1994: Drought Follows the Plow. Cambridge University Press, 175 pp.

  • Gregory, J. M., , J. F. B. Mitchell, , and A. J. Brady, 1997: Summer drought in northern midlatitudes in a time-dependent CO2 climate experiment. J. Climate, 10 , 662686.

    • Search Google Scholar
    • Export Citation
  • Groisman, P. Y., , and D. R. Easterling, 1994: Variability and trends of total precipitation and snowfall over the United States and Canada. J. Climate, 7 , 184205.

    • Search Google Scholar
    • Export Citation
  • Groisman, P. Y., , and D. R. Legates, 1994: The accuracy of United States precipitation data. Bull. Amer. Meteor. Soc., 75 , 215227.

  • Groisman, P. Y., , R. W. Knight, , and T. R. Karl, 2001: Heavy precipitation and high streamflow in the contiguous United States: Trends in the twentieth century. Bull. Amer. Meteor. Soc., 82 , 219246.

    • Search Google Scholar
    • Export Citation
  • Guttman, N. B., , and R. G. Quayle, 1996: A historical perspective of U.S. climate divisions. Bull. Amer. Meteor. Soc., 77 , 293303.

  • Halpert, M. S., , and C. F. Ropelewski, 1992: Surface temperature patterns associated with the Southern Oscillation. J. Climate, 5 , 577593.

    • Search Google Scholar
    • Export Citation
  • Hollander, M., , and D. A. Wolfe, 1999: Nonparametric Statistical Methods. 2d ed. Wiley and Sons, 787 pp.

  • Hu, Q., , C. M. Woodruff, , and S. E. Mudrick, 1998: Interdecadal variations of annual precipitation in the central United States. Bull. Amer. Meteor. Soc., 79 , 221229.

    • Search Google Scholar
    • Export Citation
  • Hurrell, J. W., 1995: Decadal trends in the North Atlantic Oscillation and relationships to regional temperature and precipitation. Science, 269 , 676679.

    • Search Google Scholar
    • Export Citation
  • Jones, P. D., cited. 2001: Gridded hemispheric and global temperature averages. [Available online at http://www.cru.uea.ac.uk/cru/data/temperature.].

    • Search Google Scholar
    • Export Citation
  • Kaplan, A., , M. Cane, , Y. Kushnir, , A. Clement, , M. Blumenthal, , and B. Rajagopalan, 1998: Analyses of global sea surface temperature 1856–1991. J. Geophys. Res, 103 , 1856718589.

    • Search Google Scholar
    • Export Citation
  • Karl, T. R., , and P. D. Jones, 1989: Urban bias in area-averaged surface temperature trends. Bull. Amer. Meteor. Soc., 70 , 265270.

  • Karl, T. R., , and R. W. Knight, 1998: Secular trends of precipitation amount, frequency, and intensity in the USA. Bull. Amer. Meteor. Soc., 79 , 231241.

    • Search Google Scholar
    • Export Citation
  • Karl, T. R., , L. Metcalf, , M. L. Nicodemus, , and R. Quayle, 1983: Statewide average climatic history. Historical Climatology Series 6-1, National Climate Data Center, 35 pp.

    • Search Google Scholar
    • Export Citation
  • Karl, T. R., , C. N. Williams Jr., , P. J. Young, , and W. Wendland, 1986: A model to estimate the time of observation bias associated with monthly mean maximum, minimum and mean temperatures for the United States. J. Climate Appl. Meteor., 25 , 145160.

    • Search Google Scholar
    • Export Citation
  • Karl, T. R., and Coauthors. 1993: A new perspective on recent global warming: Asymmetric trends in daily maximum and minimum temperature. Bull. Amer. Meteor. Soc., 74 , 10071023.

    • Search Google Scholar
    • Export Citation
  • Karl, T. R., , R. W. Knight, , and N. Plummer, 1995: Trends in high frequency climate variability during the 20th century. Nature, 377 , 217220.

    • Search Google Scholar
    • Export Citation
  • Karl, T. R., , R. W. Knight, , D. R. Easterling, , and R. G. Quayle, 1996: Indices of climate change for the United States. Bull. Amer. Meteor. Soc., 77 , 279292.

    • Search Google Scholar
    • Export Citation
  • Kiladis, G. N., , and H. F. Diaz, 1989: Global climatic anomalies associated with extremes in the Southern Oscillation. J. Climate, 2 , 10691090.

    • Search Google Scholar
    • Export Citation
  • Kunkel, K. E., , K. Andsager, , and D. R. Easterling, 1999: Long-term trends in extreme precipitation events over the conterminous United States and Canada. J. Climate, 12 , 25152527.

    • Search Google Scholar
    • Export Citation
  • Langbein, W. B., , and J. R. Slack, 1982: Yearly variations in runoff and frequency of dry years for the conterminous United States. U.S. Geological Survey Open File Rep. 82-751, 85 pp.

    • Search Google Scholar
    • Export Citation
  • Lau, K. M., , and H. Weng, 1995: Climate signal detection using wavelet transform: How to make a time series sing. Bull. Amer. Meteor. Soc., 76 , 23912402.

    • Search Google Scholar
    • Export Citation
  • Legates, D. R., 1995: Precipitation measurement biases and climate change detection. Preprints, Sixth Symp. on Global Change Studies, Dallas, TX, Amer. Meteor. Soc., 168–173.

    • Search Google Scholar
    • Export Citation
  • Lettenmaier, D. P., , E. F. Wood, , and J. R. Wallis, 1994: Hydro-climatological trends in the continental United States, 1948–88. J. Climate, 7 , 586607.

    • Search Google Scholar
    • Export Citation
  • Lins, H. F., , and J. R. Slack, 1999: Streamflow trends in the United States. Geophys. Res. Lett., 26 , 227230.

  • Livezey, R. E., , M. Masutani, , A. Leetma, , H. Rui, , M. Ji, , and A. Kumar, 1997: Teleconnective response of the Pacific–North American region to large central equatorial Pacific SST anomalies. J. Climate, 10 , 17871820.

    • Search Google Scholar
    • Export Citation
  • Lockeretz, W., 1978: The lessons of the dust bowl. Amer. Sci., 66 , 560569.

  • Manabe, S., , and R. T. Wetherald, 1987: Large-scale changes of soil wetness induced by an increase in atmospheric carbon dioxide. J. Atmos. Sci., 44 , 12111235.

    • Search Google Scholar
    • Export Citation
  • Mann, H. B., , and D. R. Whitney, 1947: On a test of whether one of two random variables is stochastically larger than the other. Ann. Math. Stat., 18 , 5060.

    • Search Google Scholar
    • Export Citation
  • Mantua, N. J., , S. R. Hare, , Y. Zhang, , J. M. Wallace, , and R. C. Francis, 1997: A Pacific interdecadal climate oscillation with impacts on salmon production. Bull. Amer. Meteor. Soc., 78 , 10691079.

    • Search Google Scholar
    • Export Citation
  • Mauget, S. A., , and D. R. Upchurch, 1999: El Niño and La Niña related climate and agricultural impacts over the Great Plains and Midwest. J. Prod. Agric., 12 , 203214.

    • Search Google Scholar
    • Export Citation
  • Mendenhall, W., , D. D. Wackerly, , and R. L. Sheaffer, 1990: Mathematical Statistics with Applications. 4th ed. PWS-Kent, 788 pp.

  • Mote, T., 1996: Influence of ENSO on maximum, minimum, and mean temperatures in the southeast United States. Phys. Geogr., 17 , 497512.

    • Search Google Scholar
    • Export Citation
  • Namias, J., 1966: Nature and possible causes of the northeastern United States drought during 1962–1965. Mon. Wea. Rev., 94 , 543554.

    • Search Google Scholar
    • Export Citation
  • Namias, J., 1982: Anatomy of Great Plains protracted heat waves (especially the 1980 U.S. summer drought). Mon. Wea. Rev., 110 , 824838.

    • Search Google Scholar
    • Export Citation
  • Namias, J., 1989: Cold waters and hot summers. Nature, 338 , 1516.

  • Namias, J., 1991: Spring and summer drought over the contiguous U.S.—Causes and prediction. J. Climate, 4 , 5465.

  • Namias, J., , X. Yuan, , and D. R. Cayan, 1988: Persistence of North Pacific sea surface temperature and atmospheric flow patterns. J. Climate, 1 , 682703.

    • Search Google Scholar
    • Export Citation
  • National Academy of Sciences, 1975: Understanding Climatic Change: A Program for Action. National Academy of Sciences, 239 pp.

  • National Research Council, 2002: Abrupt Climate Change: Inevitable Surprises. National Academy Press, 244 pp.

  • Oglesby, R. J., , and D. J. Erickson, 1989: Soil moisture and the persistence of North American drought. J. Climate, 2 , 13621380.

  • Palmer, T. N., , and C. Brankoviĉ, 1989: The 1988 U.S. drought linked to anomalous sea surface temperature. Nature, 338 , 5457.

  • Parker, D. E., , P. D. Jones, , C. K. Folland, , and A. Bevan, 1994: Interdecadal changes of surface temperature since the late 19th century. J. Geophys. Res., 99 , 1437314399.

    • Search Google Scholar
    • Export Citation
  • Ponte, L., 1976: The Cooling. Prentice-Hall, 268 pp.

  • Puckett, W. E., , J. H. Dane, , and B. F. Hajek, 1985: Physical and mineralogical data to determine soil hydraulic properties. Soil Sci. Soc. Amer. J., 49 , 831836.

    • Search Google Scholar
    • Export Citation
  • Quayle, R. G., , D. R. Easterling, , T. R. Karl, , and P. J. Hughes, 1991: Effects of recent thermometer changes in the cooperative station network. Bull. Amer. Meteor. Soc., 72 , 17181723.

    • Search Google Scholar
    • Export Citation
  • Rind, D., 1982: The influence of ground moisture conditions in North America on summer climate as modeled in the GISS GCM. Mon. Wea. Rev., 110 , 14871493.

    • Search Google Scholar
    • Export Citation
  • Rind, D., , R. Goldberg, , J. Hansen, , C. Rosenzweig, , and R. Ruedy, 1990: Potential evaporation and the likelihood of future drought. J. Geophys. Res., 95 , (D7),. 998310004.

    • Search Google Scholar
    • Export Citation
  • Rogers, J., 1990: Patterns of low-frequency monthly sea level pressure variability (1899–1986) and associated wave cyclone frequencies. J. Climate, 3 , 13641379.

    • Search Google Scholar
    • Export Citation
  • Ropelewski, C. F., , and M. S. Halpert, 1989: Precipitation patterns associated with the high index phase of the Southern Oscillation. J. Climate, 2 , 268283.

    • Search Google Scholar
    • Export Citation
  • Rosenzweig, C., , and D. Hillel, 1993: The Dust Bowl of the 1930's: Analog of greenhouse effect in the Great Plains? J. Environ. Qual., 22 , 922.

    • Search Google Scholar
    • Export Citation
  • Schaal, L. A., , and R. F. Dale, 1977: Time of observation temperature bias and “climatic change.”. J. Appl. Meteor., 16 , 215222.

  • Shabbar, A., , and M. Khandekar, 1996: The impact of El Niño–Southern Oscillation on the temperature field over Canada. Atmos.–Ocean, 34 , 401416.

    • Search Google Scholar
    • Export Citation
  • Shukla, J., , and Y. Mintz, 1982: Influence of land-surface evapotranspiration on the Earth's climate. Science, 215 , 14981500.

  • Stephenson, D. B., cited. 1999: The North Atlantic Oscillation thematic website. [Available online at http://www.met.rdg.ac.uk/cag/NAO/.].

    • Search Google Scholar
    • Export Citation
  • Tannehill, I. R., 1947: Drought, Its Causes and Effects. Princeton University Press, 264 pp.

  • Thiebaux, H. J., , and F. W. Zwiers, 1984: The interpretation and estimation of effective sample size. J. Climate Appl. Meteor., 23 , 800811.

    • Search Google Scholar
    • Export Citation
  • Trenberth, K. E., 1990: Recent observed climate changes in the Northern Hemisphere. Bull. Amer. Meteor. Soc., 71 , 988993.

  • Trenberth, K. E., , and G. W. Branstator, 1992: Issues in establishing causes of the 1988 drought over North America. J. Climate, 5 , 159172.

    • Search Google Scholar
    • Export Citation
  • Trenberth, K. E., , and J. W. Hurrell, 1994: Decadal atmosphere–ocean variations in the Pacific. Climate Dyn., 9 , 303319.

  • Trenberth, K. E., , G. W. Branstator, , and P. A. Arkin, 1988: Origins of the 1988 North American Drought. Science, 242 , 16401645.

  • Wessel, P., , and W. H. F. Smith, 1995: New version of the Generic Mapping Tools released. Eos, Trans. Amer. Geophys. Union, 76 , 329.

  • Wilcoxon, F., 1945: Individual comparisons by ranking methods. Biometrics Bull., 1 , 8083.

  • Wilks, D. S., 1995: Statistical Methods in the Atmospheric Sciences. Academic Press, 464 pp.

  • Wolfson, N., , R. Atlas, , and Y. C. Sud, 1987: Numerical experiments related to the summer 1980 heat wave. Mon. Wea. Rev., 115 , 13451357.

    • Search Google Scholar
    • Export Citation
  • Woodhouse, C. A., , and J. T. Overpeck, 1998: 2000 years of drought variability over the central United States. Bull. Amer. Meteor. Soc., 79 , 11451162.

    • Search Google Scholar
    • Export Citation
  • Worster, D., 1979: The Dust Bowl: The Southern Plains in the 1930's. Oxford University Press, 277 pp.

  • Zhang, Y., , J. M. Wallace, , and D. S. Battisti, 1997: ENSO-like interdecadal variability: 1900–93. J. Climate, 10 , 10041020.

  • View in gallery

    (a) Time series of cumulative Nov–Oct rainfall spatially averaged over the 343 climate divisions of the coterminous United States (NPCP) for 1896–1999. Black bars indicate the 10 most highly ranked water years. Water years are identified with calendar year containing the final ten months of each Nov–Oct period; i.e., 1896 refers to Nov 1895 to Oct 1896, etc. (b) Mann–Whitney Z statistics for running 27-yr samples of NPCP rankings. Horizontal lines indicate positive and negative significance at 90%, 95%, and 99% confidence levels

  • View in gallery

    (a) Percentage of U.S. area showing significant (>1.645) Mann–Whitney Z (MWZ) statistics for Nov–Oct rainfall over running 27-yr time windows during 1932–99, with rankings calculated over the 1932–99 base period. (b) MWZ statistics for annual rainfall rankings at the climate division level during 1972–98. Gray shaded and hatched climate divisions show statistics significant at 90%, 95%, and 99% confidence levels

  • View in gallery

    (a) N-year time windows during which relatively large areas of the United States experienced a significant incidence of high (MWZ > 1.645) rankings in annual rainfall, as determined by timelines similar to Fig. 2a. Black bars show peak wet periods at each sample size for N = 6 to 30 yr. Gray bars mark secondary peak periods, defined here as periods during which significant Z statistics were found over areas comparable to peak area values, but not overlapping the primary peak period in time. (b) Percentage of total U.S. area showing significance at a 90% or better confidence level during the corresponding peak wet periods in (a). (c) As in (a) for peak and secondary periods marked by a significant incidence of low (i.e., MWZ < −1.645) rankings in annual rainfall. (d) Percentage of U.S. area showing significant MWZ statistics during corresponding peak dry periods in (c)

  • View in gallery

    (a) As in Fig. 3a for N-year time windows marked by a significant incidence of high rankings in annual temperature. (b) Percentage of total U.S. area showing significance at a 90% or better confidence level during the corresponding peak warm periods in (a). (c) As in (a) for peak and secondary periods marked by a significant incidence of low rankings in annual temperature. (d) Percentage of U.S. area showing significant MWZ statistics during corresponding peak cool periods in (c)

  • View in gallery

    (a) Timelines of significant dry, wet, cool, and warm periods as indicated by Figs. 3 and 4. (b) 1920–99 monthly cold tongue SSTA anomalies as derived from Kaplan et al. (1998) analysis values and regression analysis. (c) Composite PDO index formed from unfiltered and normalized SSTA analysis values averaged over the PDO1 and PDO2 regions outlined in (e). (d) 1920–99 annual mean Northern Hemisphere surface temperature anomaly (Parker et al. 1994). (e) First unrotated EOF of low-frequency (υ < 72 months−1) northern Pacific SSTA analyses over 1920–91, and cold tongue region as defined here

  • View in gallery

    (a) As in Fig. 2b for annual rainfall rankings during 1933–40. (b) As in Fig. 2b for annual temperature rankings during 1932–39. (c) Increase in acreage of harvested crops between 1919 and 1929 (from Tannehill 1947)

  • View in gallery

    (a) As in Fig. 2b for annual rainfall rankings during 1950–56. (b) As in Fig. 2b for annual temperature rankings during 1949–56

  • View in gallery

    As in Fig. 2b for annual temperature rankings during 1964–79

  • View in gallery

    (a) As in Fig. 2b for annual rainfall rankings during 1982–99. (b) As in Fig. 2b for annual temperature rankings during 1986–99. (c) As in Fig. 2b for annual temperature rankings during 1986–92

  • View in gallery

    (a) As in Fig. 2b for annual rainfall rankings during 1963–69. (b) As in Fig. 2b for annual rainfall rankings during 1972–79. (c) Northern winter (DJFM) North Atlantic Oscillation values for 1920–98 (Hurrell 1995).

  • View in gallery

    Trend variation in annual temperatures during 1957–99. Temperature values indicated are trend coefficients derived from linear trend analysis of temperature during 1957–99, multiplied times 43 (yr)

All Time Past Year Past 30 Days
Abstract Views 0 0 0
Full Text Views 46 46 6
PDF Downloads 13 13 0

Intra- to Multidecadal Climate Variability over the Continental United States: 1932–99

View More View Less
  • 1 U.S. Department of Agriculture–Agricultural Research Service, Lubbock, Texas
© Get Permissions
Full access

Abstract

Trend analysis is used frequently in climate studies, but it is vulnerable to a number of conceptual shortcomings. This analysis of U.S. climate division data uses an alternate approach. The method used here subjects time series of annual average temperature and total precipitation to tests of Mann–Whitney U statistics over moving sampling windows of intra- to multidecadal (IMD) duration. In applying this method to time series of nationally averaged annual rainfall, a highly significant incidence of wet years is found after the early 1970s. When applied to individual climate divisions this test provides the basis for a climate survey method that is more robust than linear trend analysis, and capable of objectively isolating the timing and location of major IMD climate events over the United States. From this survey, four such periods emerge between 1932 and 1999: the droughts of the 1930s and 1950s, a cool 1964–79 period, and wet–warm time windows at the end of the century. More circumstantial consideration was also given here to the state of ENSO, the Pacific decadal oscillation (PDO), the winter state of the North Atlantic Oscillation, and mean annual Northern Hemisphere surface temperature during those periods. Anecdotal evidence presented here suggests that wet years associated with warm-phase ENSO conditions and the positive phase of the PDO may have played a role in ending the drought periods of the 1930s and 1950s. Conversely, the La Niña–like climate impacts found here during the late 1940s to mid-1950s, and the increased incidence of cold phase ENSO and negative phase PDO conditions during that time, suggests connections between that ocean state and severe drought. Significant late-century warmth was found mainly in the western United States after the mid-1980s, but no evidence of a cooling trend was evident in the southeast, as reported elsewhere. The late-century wet regime appears to have occurred in two phases, with wetness confined to the east during 1972–79, and more concentrated in the southwest and central United States during 1982–99.

Corresponding author address: Steven A. Mauget, USDA Plant Stress and Water Conservation Laboratory, 3810 4th Street, Lubbock, TX 79415. Email: smauget@lbk.ars.usda.gov

Abstract

Trend analysis is used frequently in climate studies, but it is vulnerable to a number of conceptual shortcomings. This analysis of U.S. climate division data uses an alternate approach. The method used here subjects time series of annual average temperature and total precipitation to tests of Mann–Whitney U statistics over moving sampling windows of intra- to multidecadal (IMD) duration. In applying this method to time series of nationally averaged annual rainfall, a highly significant incidence of wet years is found after the early 1970s. When applied to individual climate divisions this test provides the basis for a climate survey method that is more robust than linear trend analysis, and capable of objectively isolating the timing and location of major IMD climate events over the United States. From this survey, four such periods emerge between 1932 and 1999: the droughts of the 1930s and 1950s, a cool 1964–79 period, and wet–warm time windows at the end of the century. More circumstantial consideration was also given here to the state of ENSO, the Pacific decadal oscillation (PDO), the winter state of the North Atlantic Oscillation, and mean annual Northern Hemisphere surface temperature during those periods. Anecdotal evidence presented here suggests that wet years associated with warm-phase ENSO conditions and the positive phase of the PDO may have played a role in ending the drought periods of the 1930s and 1950s. Conversely, the La Niña–like climate impacts found here during the late 1940s to mid-1950s, and the increased incidence of cold phase ENSO and negative phase PDO conditions during that time, suggests connections between that ocean state and severe drought. Significant late-century warmth was found mainly in the western United States after the mid-1980s, but no evidence of a cooling trend was evident in the southeast, as reported elsewhere. The late-century wet regime appears to have occurred in two phases, with wetness confined to the east during 1972–79, and more concentrated in the southwest and central United States during 1982–99.

Corresponding author address: Steven A. Mauget, USDA Plant Stress and Water Conservation Laboratory, 3810 4th Street, Lubbock, TX 79415. Email: smauget@lbk.ars.usda.gov

1. Introduction

Many have relied on trend analysis to identify climate change over the United States and North America in multidecadal or centennial timescale data records (e.g., Lettenmaier et al. 1994; Karl et al. 1993, 1995, 1996; Karl and Knight 1998; Groisman and Easterling 1994; Kunkel et al. 1999; Groisman et al. 2001). However, climate variation that projects significantly on a long-term linear trend may constitute a relatively special case. In some cases low-frequency variation during the time intervals over which trends are fitted may be more cyclic in nature. For example, some regions in the United States experienced warm conditions during the 1930s and 1950s, followed by a relative cooling during the 1960s and 1970s. In such areas a return to warm conditions at the end of the century might result in a weak trend in a data record beginning in 1932, even though that late warming may mark a significant climate shift. Conversely, a data record beginning in 1960 might result in a strongly significant temperature trend, thus the significance or even the sign of trends may be sensitive to the choice of period over which trends are fitted. This shortcoming might be partially overcome by trend fitting over a range of time windows, as in Lins and Slack's (1999) analysis of 30-, 40-, 50-, 60-, 70-, and 80-yr streamflow records ending in 1993. In other cases where significant trends are found, that significance may not indicate a systematic climate shift. Thus rainfall trends over the central United States that experienced wet conditions over the past 20 years and dryness during the drought years of the 1930s might be significantly positive, but the significance of that trend would be partly due to the unrelated climate regimes near the data record's beginning and end. An example of this can be found in Kunkel et al.'s (1999) trend analysis of 1–7-day extreme precipitation events over the United States between 1931 and 1996. While they note that the national trend for these events is significantly positive, they also note that significance is partially a result of the low-frequency dryness of the 1930s and 1950s. While these are examples of potential problems, the fundamental problem is that whereas climate variability may be abruptly transitional (National Research Council 2002) or quasi-cyclic, trend analysis assumes somewhat linear behavior and translates nonlinear variability into one trend value. As will be seen here, the U.S. climate between 1932 and 1999 was marked by intradecadal and longer-term variation, some of which was inconsistent with gradual and monotonic change.

To avoid the potential problems associated with trend analysis the approach here will be to test relatively simple expressions of U.S. climate—annual rainfall and mean annual temperature—using nonparametric statistics derived from running samples of those two variables rankings. The result is a moving-window analysis method similar in concept to wavelet analysis (e.g., Lau and Weng 1995), but substituting the evaluation of Mann–Whitney U statistics for that of a wavelet transform. A primary goal here is to objectively test for evidence of relatively recent climate change by testing multidecadal climate records for significant impacts over a range of intra- to multidecadal (IMD) timescales. As a result, an underlying assumption made here is that climate change over the United States might be best detected as a special case of IMD climate variation. But in addition, such an approach also allows for establishing context in two ways: first, by considering the circumstances of more recent variability against that of earlier IMD climate events; second, by considering the circumstances of those earlier events in the context of more recently recognized climate mechanisms and variability. Section 2 of this paper will describe the data used in this analysis of U.S. climate and Pacific sea surface temperatures. Section 3 will describe the statistical tests and Monte Carlo protocol used here, and a method by which annual temperature and precipitation variability is surveyed over a range of IMD timescales. Section 4 will narrow the focus to four major IMD climate periods over the United States during 1932–99, and will consider those periods in the context of four indicators of Pacific, Atlantic, and Northern Hemisphere climate introduced in section 3c. Section 5 will provide summary and concluding remarks.

2. Data

a. Climate division data

Climate division data (Guttman and Quayle 1996) is derived from averages of monthly station data reported over the continental United States's 343 climate divisions, and, as used here, extends between January 1895 and December 1999. After 1930 divisional data values were derived via equal-weight averaging of monthly average temperature or cumulative precipitation reported from stations within each climate division. Before 1931 station data was not averaged over climate divisions, but was averaged statewide or regionally by the U.S. Department of Agriculture (USDA). Current pre-1931 climate division data values are inferred from those pre-1931 USDA averages and regression relationships between climate division and statewide averages derived over the period 1931–82. Guttman and Quayle (1996) report that although correlations between regression-derived and actual divisional values during 1931–82 are high (>0.90), the variance of divisional regression estimates before 1931 are generally less than that of divisional values after 1930 that were directly averaged from station data.

There are a number of advantages in the use of climate division data in climate analysis. The data's continuity in time makes it suitable for studies that require serially complete data. Here, the fact that divisional averages are assigned to the division's area will be exploited in two ways. First, in the formation of national spatial averages of annual climate variables, as in Karl et al. (1996), and also to estimate the percentage of the continental U.S. area that shows significant climate impacts over specified time windows. The most useful quality of this data might be traced to the relatively dense sampling of the underlying station network. Averaging an adequate sample of stations can result in the cancellation of random measurement error, although systematic bias such as winter precipitation undercatch (Groisman and Legates 1994; Groisman and Easterling 1994) would remain. But the results derived here are, in principle, insensitive to measurement biases that are constant over time, as the associated methods are based on the analysis of rankings of annual climate variables.

But there are also potential problems with the use of climate division data. The primary concern here are observational biases that vary over time, particularly those that are widely introduced into the observing network. In addition to the 1931 variance discontinuity mentioned above, the number and locations of stations within a divisional region can vary over the period of record. This changing network composition can lead to artificial inhomogeneities, particularly over climate divisions spanning different climate regimes. Users are specifically warned (Karl et al. 1983) that the spatial distribution in mountainous western states was not uniform in time before 1931. As a result, a time-varying bias is possible in early state averages and the pre-1931 divisional values derived from them, although a correction procedure is applied to adjust for that bias (Karl et al. 1983; Guttman and Quayle 1996).

In temperature measurements a gradual shift to morning observing times has been found to produce an artificial cooling bias in station data (Schaal and Dale 1977), which has been corrected for in climate division temperature averages (Karl et al. 1986). The relatively rapid conversion to the maximum–minimum temperature system during the 1980s involved about 3300 of the United States's 7750 cooperative stations (Quayle et al. 1991). This conversion resulted in a cool bias in mean monthly temperatures, which, given the number of stations involved, could produce a similar shift in climate division averages. This is believed to have played a minor role here, as significant warming is indicated after 1980 in spite of a possible cooling bias in the observing network. However, it is possible that such a bias may have reduced the significance of those results. Although urban warming (Karl and Jones 1989) is a potential problem in climate divisions with developing urban settings, it is unlikely to have significantly influenced the results presented here, as the warming mentioned above is evident over some of the United States's least populated regions.

The relocation of most first-order weather stations to suburban airports during the 1930s through the 1950s, and the introduction of Alter wind shields to reduce precipitation undercatch at many of those stations (Groisman and Legates 1994; Groisman and Easterling 1994), could introduce artificial shifts into precipitation totals. Yet these changes may have a minor effect here, as first-order stations account for a small fraction of those contributing to climate division averages. A more plausible bias could result from a widespread shift from snowfall to rainfall during the winter months. Given the evidence of late-century warming found here, such a shift would improve the meteorological network's ability to detect winter precipitation, even if total precipitation and the observing network itself were unchanged (Legates 1995). This is a potential but unresolved source of bias here, and a possible alternate explanation for the significant late-twentieth-century wetness found here.

b. Kaplan historical sea surface temperature analysis

In forming Pacific sea surface temperature anomaly (SSTA) indices the optimally smoothed version of the reduced space historic SSTA analysis of Kaplan et al. (1998) was used here. Their analyzed SSTA values are defined on a 5° × 5° grid at monthly time resolution and are continuous in time over the tropical and North Pacific from January 1856 to December 1991. Kaplan analysis values are reported as anomalies calculated with respect to a 1951–80 climatology.

3. Methods

a. Monte Carlo methods and Mann–Whitney U and Z statistics

The statistical methods used here, that is, the derivation of Mann–Whitney U and Z statistics and a Monte Carlo protocol, will be described through an analysis of spatially averaged annual rainfall over the continental United States.

Assuming a November–October water year, 104 yr of annual precipitation (apcp) totals can be derived for each climate division over the period 1895–1999. Using those yearly totals and the areas of each climate division, a national spatial average of each year's precipitation (NPCP) can be estimated:
i1520-0442-16-13-2215-e1
An interesting feature of this record (Fig. 1a) is that a high incidence of wet years are apparent over the 27 years between 1973 and 1999. In fact, when all 104 annual averages are ranked, it is found that 9 of the 10 wettest years1 occurred during that period. If the statistics of U.S. rainfall were essentially stationary over the past century those wet years would be more randomly distributed throughout the 1896–1999 water years.2 Assuming such a null hypothesis, the probability of finding this incidence of wet years in a 27-yr sample drawn from a stationary 104-yr population could be estimated using a hypergeometric probability distribution (Mendenhall et al. 1990). However, classical hypergeometric statistics assume independent data sampling, while IMD climate variation will reduce the number of temporal degrees of freedom in a climate record of annual values (Thiebaux and Zwiers 1984). A more accurate estimate of the significance of the incidence of wet years in Fig. 1a could be derived using a Monte Carlo generated null distribution. Such distributions were formed here via the following Monte Carlo protocol, which—with minor variations—is used throughout this research.
  1. Form red noise models consistent with a null hypothesis that data time series represent a stochastic climate sharing the same mean and variance of the actual climate record, and similar persistence characteristics, but are essentially trendless and stationary over the period of analysis (H0). The initial stage in this model formulation process calculates the AR(1), AR(2), and AR(3) regression coefficients from the autocorrelation values of the detrended data, and selects the AR model yielding the minimum Akaike information criteria score (Akaike 1974).
  2. From the results of step 1, form autoregressive red noise processes.
  3. Adjust the mean and variance of the red noise process resulting from step 2 and truncate the number of significant digits to agree with that of the data. Then, select red noise series of appropriate length—in this case, 104—and rank those values.
  4. From the ranked noise processes resulting from step 3 calculate appropriate null statistics, which in this specific case is the number of the 10 most highly ranked values found in nonoverlapping 27-element segments of each red noise series.
  5. Repeat steps 2–4 until 10 000 independent null realizations are calculated, and determine the distribution or the distribution parameters of the resulting null statistics.

In 10 separate repetitions of this protocol, 9 or more “top-ten” values were found on 8 occasions; that is, 8 times out of 100 000 null realizations. As a result, the incidence of top-ten ranked wet years during 1973–99 is conservatively estimated here to be significant at a 99.9% confidence level. Karl et al. (1996) also note a tendency for average annual U.S. rainfall to be above the twentieth-century mean after 1970, but only assign an approximate 90% confidence level to that shift. The incidence of highly ranked post-1972 NPCP values found here confirms that tendency, but suggests a much more significant regime shift in national rainfall conditions.

The above hypergeometric test was based on a specific but arbitrary rule for marking wet years, that is, the incidence of the 10 most highly ranked years in a 27-yr sample. Although top-ten tests and lists are easily understood, this particular test lacks in generality. The Mann–Whitney U Statistic (Wilcoxon 1945; Mann and Whitney 1947; Mendenhall et al. 1990) can be used to determine the significance of an arbitrary distribution of rankings in a sample. As used here, the formation of Mann–Whitney U statistics assumes the data have been ranked and divided into two classes. For example, the 104 NPCP values of Fig. 1a might be divided into two classes of nI and nII elements: the 77 yr between 1896 and 1972 (class I), and the 27 yr between 1973 and 1999 (class II). Although the analytic form in Wilks (1995) was used here, the U statistic for class II is equivalent to the total number of class I members that precede each member of class II when all data values are arranged by rank (Hollander and Wolfe 1999). That is,
i1520-0442-16-13-2215-e2
where Rank Ii is the rank of the ith member of class I, etc., and φ(Rank Ii, Rank IIj) = 1 if Rank Ii < Rank IIj, 0 otherwise. Thus the maximum UII statistic in this example would occur when that class accounts for the 27 highest rankings (UII = 77 × 27), while the smallest statistic would result when class II accounts for the 27 lowest (UII = 0 × 27). The 104!/77!27! arrangements that class II rankings might assume produce normally distributed U statistics with a mean equal to the average of the maximum and minimum values [e.g., μU = 0.5(77 × 27)], and a standard deviation given by the expression σU = [nInII(nI + nII + 1)/12]1/2. These parameters can be used to Z-transform U statistics, with significantly high (low) Z statistics indicating a significant incidence of high (low) rankings in a sample relative to a null hypothesis that assumes random sampling. However, the desire here is to compare observed U statistics against a more climate-specific null hypothesis similar to that posed in the earlier hypergeometric test, that is, that climate variation consists of a stochastic but persistent process that is essentially stationary in the long term. Thus U null statistics were derived via Monte Carlo simulations identical to those used to derive hypergeometric null statistics as described in steps 1 through 5 above, but with U statistics calculated over nonoverlapping sampling windows in step 4. The mean (μMC) and standard deviation (σMC) of the resulting 10 000 U null values were then used to normalize U statistics derived from rankings sampled from the undetrended time series of annual data. As a result, the general incidence of observed data rankings can be tested against the hypothesis of stationarity posed in the Monte Carlo protocol (H0):
i1520-0442-16-13-2215-e3
The resulting Mann–Whitney Z (MWZ) statistic for all 1973–99 NPCP rankings (2.94) is significant at better than a 99.0% confidence level. To put that result into context, the results of running analyses of MWZ statistics for that and all other possible 27-yr NPCP samples during 1896–1999 can be found in Fig. 1b. In that figure, only the 1972–98 MWZ statistic (3.03) exceeds the 1973–99 value.

Because of questions regarding how representative climate division data are before 1931, the hypergeometric and Mann–Whitney analyses of Fig. 1a were repeated using NPCP values ranked over the period 1932–99. General rankings change over that period although the top-ten years remain the same, having all occurred after 1931. As found in Table 1, the estimated significance levels for the incidence of top-ten wet years after 1972 is somewhat lower due to the reduced population size. Peak MWZ values also decrease, but occur over the same 27-yr time window and remain significant at at least a 99.0% confidence level. To test for sensitivity to the definition of water year, the 1932–99 analyses were also repeated with NPCP values accumulated between October and September. Nine of the 10 wettest October–September water years during that 68-yr period also occurred after 1972, although the lone pre-1973 top-ten year in the November–October analysis, 1941, was replaced by 1958 in the October–September analysis. In the following climate survey, monthly climate variables at the climate division level will be integrated or averaged over November–October periods, given the desire to fully capture both the midwinter rainfall maxima over the West Coast and the temperature minima over the country as a whole. Also, analysis will be limited to the years 1932–99 because of reservations about pre-1931 data, particularly in western climate divisions.

b. Intra- to multidecadal climate survey

In Fig. 1 annual rainfall was spatially averaged over all climate divisions, and the resulting rankings from running 27-yr time windows were then subjected to MWZ analysis. In developing a method to survey both the timing and location of significant IMD climate variability, that process is essentially reversed. That is, MWZ analysis of time series of ranked annual data from individual climate divisions is first conducted, then the percentage of U.S. area that shows significant results during running time windows is noted. The method used to evaluate samples of rankings at the climate division level is identical to that used to test U statistics derived from running 27-yr samples of NPCP rankings. Null U distribution parameters are determined via the five-step simulation process described before, with μMC and σMC parameters derived for both annual temperature and rainfall series at each climate division. As in Fig. 1b, the calculation of division-level MWZ statistics were repeated over running sampling windows during 1932–99 and the percentage of U.S. area showing both positive and negative significance during each sampling period was recorded.

An example of this approach can be found in Figs. 2a and 2b. Figure 2a shows the area percentages that result when annual rainfall rankings over all 27-yr sampling windows during 1932–99 are tested for positive significance (MWZ > 1.645). As in Fig. 1b, the 1972–98 result is the maximum value, with approximately 40% of U.S. area showing a significant incidence of highly ranked annual rainfall. In Fig. 2b gray shaded climate divisions mark those whose MWZ statistics for 1972–98 annual rainfall rankings are positively significant at a 90% level or better, with darker shades showing higher significance. In that figure significant positive MWZ statistics are the norm. Conversely, significant negative values, as indicated by hatched climate divisions, are the exception.

Mann–Whitney Z statistics were introduced here as a more general statistical test, yet so far this test has been limited to running 27-yr time windows. To screen for climate variation over a range of timescales MWZ tests were repeated here using sampling windows of varying duration. Thus analyses similar to that leading to Fig. 2a were conducted with 25 sample window sizes between 6 and 30 yr in length. As before, the Monte Carlo trials leading to Eq. (3)'s μMC and σMC parameters were conducted for both annual variables at each climate division. In addition, varying window duration required Monte Carlo trials to be repeated for each window length, given the dependence of μU and σU on sample size. Thus for each climate division, for both annual rainfall and annual mean temperature, 25 μMCσMC parameter pairs were independently generated through the course of the Monte Carlo simulations.

For each of the 25 sample window sizes, running analyses of significant positive and negative annual temperature and rainfall Z statistics over the continental United States produced four area timelines similar to Fig. 2a, resulting in a total of 100 timelines. Again, the peak values in these timelines indicate N-year periods during which relatively large areas of the United States experienced a significant incidence of low (MWZ < −1.645) or high (MWZ > 1.645) rankings in annual rainfall or temperature. Given the number of timelines generated, only the peak periods are shown here in Figs. 3 and 4.

In Fig. 3a black bars show peak wet periods at each sample size. The gray bars in that figure mark secondary peak periods, defined here as time windows during which significant MWZ statistics were found over areas comparable to peak area values, but did not overlap the primary peak period in time. Comparable areas were considered here as at least 70% of the peak period area. Thus Figs. 3a,c and 4a,c mark IMD time windows of anomalous wet, dry, warm, and cool conditions respectively. The corresponding percentage of total U.S. area exhibiting significant MWZ statistics during each primary peak time window are found in Figs. 3b,d and 4b,d. As a result, the peak percentage for the 27-yr analysis in Fig. 2a during 1972–98 is found again in the 27-yr analysis of Fig. 3b. Figures 3 and 4 are used here as departure points to identify when and where significant IMD climate variation has occurred between 1932 and 1999. Thus, for example, the peak area values found in Figs. 3b,d and 4b,d might direct us toward the sampling window size that resulted in the most spatially extensive climate impacts, while the corresponding black and gray bars in Figs. 3a,c and 4a,c mark time windows during which those impacts occurred. Over those time windows, MWZ mappings similar to Fig. 2b can be plotted to determine where those impacts occurred.

Whereas trend analysis selects for essentially linear or monotonic behavior, the method described above screens data for a wider range of climate variability. As U statistics are calculated over all possible time windows at a given sample size, and those running analyses are repeated over a continuous range of sample sizes, this screening approach can in principle identify climate regimes of arbitrary onset and somewhat arbitrary duration (i.e., 6–30 yr) during 1932–99. The Mann–Whitney U statistic introduces an element of objectivity, in that it imposes no arbitrary thresholds defining extreme rankings but can identify a significant incidence of extreme rankings in a sample. The degree to which this process has identified the onset, duration, and distribution of more recent climate impacts might be judged by its ability to identify known instances of past IMD impacts in an unguided manner.

c. Pacific, Atlantic, and hemispheric climate variation

Given the time windows highlighted by Figs. 3 and 4, attention will be directed to four intra- to multidecadal periods (Fig. 5a): dry–warm periods of the 1930s and 1950s, the cool 1964–79 period, and wet–warm time windows after the early 1970s. In addition, the circumstances of these IMD climate events will also be considered in the context of broader climate variability during those times. That variation is expressed here in the state of four potentially related climate indices: a cold tongue SSTA index intended to gauge the state of the EI Niño–Southern Oscillation (ENSO; Fig. 5b), a composite Pacific decadal oscillation (PDO) SSTA index (Fig. 5c), average Northern Hemisphere annual surface temperature (NHT) anomalies (Fig. 5d; Parker et al. 1994), and Hurrell's (1995) northern winter North Atlantic Oscillation (NAO) index (Fig. 10c below). The general guidelines used in choosing these variables were that they exhibit IMD timescales or return times, and that they be associated with either demonstrated or potential climate impacts over the continental United States. The ENSO mechanism's effects on U.S. climate have been widely documented, and more recent work suggests those effects may depend on the state of the PDO (Gershunov and Barnett 1998; Gershunov et al. 1999). The northern winter state of the NAO varies over decadal timescales, and Rogers (1990) and Hurrel (1995) suggest its influence over the eastern United States. Hu et al. (1998) suggest the NAO's influence on decadal rainfall cycles over the central United States. The average Northern Hemisphere temperature anomaly exhibits clear low-frequency variation and is included here to compare U.S. temperature variation with broader hemispheric variability. The state of these indicators will be considered starting with the year 1920, to make the identification of climate conditions before 1932 possible.

Figure 5a's ENSO SSTA index was calculated by averaging the Kaplan SSTA analyses over the cold tongue Niño-3 region (5°S–5°N, 150°–90°W; Fig. 5e). Although the Kaplan analysis ends in December 1991, this ENSO record has been extended by linearly regressing its January 1950–December 1991 values against NCEP Niño-3.4 (5°S–5°N, 170°–120°W) values over that same period. The resulting regression coefficients and 1992–99 Niño-3.4 values were then used to derive monthly cold tongue SSTA values for 1992–99.

Figure 5c's composite PDO index was derived from the leading unrotated EOF of low-frequency (υ < 72 months−1) SSTA over a northern Pacific grid region. The first EOF of the resulting low-frequency SSTA revealed two northern Pacific centers of action, referred to here as PDO1 and PDO2 (Fig. 5e). The composite PDO index in Fig. 5c was formed from the difference of the normalized unfiltered SSTA data averaged over those regions, that is, PDO2′ − PDO1′, where PDO1′ is the unfiltered SSTA averaged over PDO1 and normalized by its 1920–91 standard deviation, etc. Between January 1920 and December 1991 the resulting PDO index is well correlated (r = 0.79) with Mantua et al.'s (1997) leading principal component of northern Pacific SSTA. As a result, the PDO index derived here was extended from January 1992 to December 1999 using their PC values for that period and 1920–91 linear regression coefficients.

Figure 5d's mean annual Northern Hemisphere surface temperature (NHT) record is the product of merged SSTA and land surface temperature anomaly datasets (Parker et al. 1994; Jones 2001). Figure 10c's winter [December–March (DJFM)] NAO index for 1920–99 is Hurrell's (1995) winter NAO index extended to 19983 and using a slightly different climatology to calculate anomalies (Stephenson 1999).

In Figs. 5b–d and 10c, thresholds marking the highest and lowest 25th percentile of values over 1920–98 or 1920–99 are included as somewhat liberal indicators of extreme conditions. In the annually resolved data these values mark the 19 extreme positive and negative anomalies in the NHT series, and the 20 extreme positive and negative index values in the winter NAO series. In the following discussion, periods of extreme positive or negative values in a given index will be referred to as by their acronym and sign, that is, ENSO(+), PDO(−), NHT(+), etc. In addition, annual rankings in the 1932–99 climate division data will be considered dry or cool if ranked 1–17, and wet or warm if ranked between 52 and 68. The terms driest, coolest, wettest, and warmest will also be used to refer to the most extreme rankings in the data, that is, 1 and 68.

4. Intra- to multidecadal climate variation: 1932–99

a. The drought of the 1930s

Figure 3c's 8-yr rainfall analysis identifies 1933–40 as a dry period, while 1932–39 is marked as an 8-yr peak warm period in Fig. 4a. The Mann–Whitney Z mappings for annual rainfall and temperature during those two time spans can be found in Figs. 6a and 6b. In Fig. 6a the core region of persistent meteorological drought straddles the 100th meridian and the 500–600-mm rainfall isohyets, boundaries traditionally used to approximate the eastern limits of the Great Plains. In Fig. 6b the most persistent temperature effects of the 1930s appear in the south and southeast. As measured by the percentage of U.S. area showing warm or warmest rankings, 1934 was the hottest year of 1932–99, with 79% of U.S. area ranked in the 52–68 range, and 38% ranked as the warmest November–October of that 68-yr period.

Obvious connections between the 1930s drought and the indices of section 3c are not clearly apparent. Aside from the ENSO(+)–PDO(+) conditions associated with the 1930/31 El Niño event, ENSO(−)–PDO(−) conditions during 1933/34, and ENSO(−) conditions during 1938/39, the 1930s appeared to be a quiet period over the tropical and North Pacific relative to the remaining years of 1932–99. While the two cold phase events may have contributed to dry conditions over the United States, it also might be argued that the absence of warm phase ENSO events may have played a role in the 1930s drought. Typically, El Niño events are associated with anomalously wet conditions during the winter months over the Great Plains (Mauget and Upchurch 1999), which supports soil moisture recharge during periods of low evaporative demand. In Fig. 5b, the longest return period between ENSO(+) conditions was the 102 months between June 1931 and January 1940. Thus warm-phase ENSO conditions, in effect, appear to have provided little relief during the 1930s.

With respect to more local influences, circumstantial evidence suggests that shifting land use patterns during the 1920s may have contributed to the persistence of drought over the Great Plains during the 1930s. Increased demand due to the reduction of Russian wheat exports to Europe during World War I, government price supports, the increased mechanization of agriculture during the 1920s, and the associated machinery debt all increased pressure to cultivate in the driest western margins of the Great Plains (Lockeretz 1978; Worster 1979). This transition is clearly shown in Fig. 6c, which shows the increase in harvested acres over those regions between 1919 and 1929. Although how such an agricultural expansion may have led to persistent meterorological drought is a matter of speculation, it seems likely that soil moisture deficits may have played a role. Such deficits may have developed through either increased evapotranspiration or through the consequences of tillage practices. Whereas 400–500 mm of annual rainfall has been cited as a threshold for farming (Glantz 1994), the average annual rainfall over the western Great Plains varies between 250 and 500 mm. As a result, the ongoing evapotranspirative demand of agriculture in such a marginal rainfall region could conceivably lead to the depletion of soil moisture. Alternatively, the loss of topsoil restricts the surface's ability to retain moisture (Puckett et al. 1985), which would have led to surface drying during the Dust Bowl years. Soil or surface moisture deficits have been associated with meteorological drought through reduced evaporation (Shukla and Mintz 1982; Rind 1982) and circulation effects on the overlying atmosphere (Wolfson et al. 1987; Oglesby and Erickson 1989). Figure 6c suggests the potential for developing such deficits from the southern Great Plains to the Canadian border during 1919–29, over a region closely coincident with the core region of meteorological drought during the 1930s. However, it is also possible that the effects of dormant vegetation (Dirmeyer 1994) and albedo (Charney et al. 1977) may have also played a role.

But anomalous climate conditions during the 1930s were not limited to the Great Plains. Widespread drought conditions over the central United States were preceded by drought in the East and Midwest in 1930, and the Great Plains dust storms were preceded by similar conditions in the Pacific Northwest in 1931 (Tannehill 1947). The results of Langbein and Slack (1982) suggest that drought conditions may have been migratory, with dryness in the west during the 1920s as a possible origin. Moreover, drought had been a recurring feature of Great Plains climate long before the introduction of agriculture (Woodhouse and Overpeck 1998). Even so, it is interesting that the most persistent and significant effects on precipitation during the 1930s appear over a marginal rainfall region that saw a widespread shift to agriculture during the previous decade. But while local effects may have contributed to the persistence of the 1930s drought over the central United States, remote effects almost certainly contributed to its end. The wet year of 1941—nationally, the fourth wettest November–October period of 1932–99—brought relief to the Great Plains and occurred during the relatively long lived ENSO(+)–PDO(+) event that spanned late 1939 to early 1942.

b. The drought of the 1950s

In Fig. 3d, 1950–56 is the most spatially extensive peak dry period over all sampling windows during 1932–99. The Mann–Whitney Z mapping for annual rainfall rankings during that 7-yr period can be found in Fig. 7a. Areas with a significantly high occurrence of dry years include New Mexico and the southern Rockies, Texas, and climate divisions along the Gulf Coast and in the southeastern states. During that time 1952, 1954, 1955, and 1956 stand out as years in which relatively large fractions of the continental United States experienced dry annual rainfall rankings. In 1956 climate divisions covering 21% of the continental United States experienced the driest November–October of 1932–99, which was the greatest percentage of area ranked 1 during any year of that period. That year also resulted in the third driest NPCP value of 1932–99. Moreover, as NPCP rankings for the years 1954 and 1955 were 5 and 8 respectively, 1956 also concluded an exceptionally dry 3-yr period over the continental United States.

In Fig. 4a 1949–56 is indicated as a peak warm period in the 8-yr temperature analysis. The MWZ mapping for annual temperature rankings for that period can be found in Fig. 7b. In that figure evidence of a significant incidence of warm years is apparent in a scattered band of climate divisions between Texas and the northeastern states. But in addition, a more coherent grouping of climate divisions along the west coast and in the northwest indicates a significant occurrence of cold annual temperatures in that region.

In contrast to the 1930s drought period, indirect evidence of external climate influences, specifically, ENSO and the PDO, seems more apparent during the 1950s. The impacts evident here in annual climate over the United States are similar to those reported for seasonal climate effects during cold-phase ENSO conditions, for example, the dry conditions found during winter and spring periods in the south and southeast (Kiladis and Diaz 1989; Ropelewski and Halpert 1989), warm winter conditions in the south and southeast (Mote 1996; Livezey et al. 1997; Mauget and Upchurch 1999), and cold winter conditions in the northern United States and Canada (Halpert and Ropelewski 1992; Shabbar and Khandekar 1996; Livezey et al. 1997). Over the northern and tropical Pacific, the period between the late 1940s and the mid-1950s saw two instances of fairly long-lived ENSO(−)–PDO(−) conditions, separated by brief periods of warm-phase ENSO conditions. In Fig. 5b a gradual shift to PDO(−) conditions is apparent beginning in early 1948, with 31 of the 38 months between March 1948 and April 1951 exhibiting PDO(−) values. That transition appears to have preceded the La Niña event of 1949/50, evident here as ENSO(−) conditions between September 1949 and December 1950. The months between early 1951 and early 1954 were marked by predominantly negative but neutral PDO values and short-lived ENSO(+) conditions during 1951 and 1953. That period was followed by a relatively long spell of ENSO(−)–PDO(−) conditions. Of the 34 months between April 1954 and January 1957, 29 exhibited ENSO(−) conditions. Roughly concurrent with those La Niña conditions was the longest sustained PDO(−) period of 1920–99, which occurred between February 1955 and January 1957. An association between cold-phase ENSO conditions and U.S. drought has been proposed before (Palmer and Branković 1989; Trenberth et al. 1988; Trenberth and Branstator 1992), but the coincidence of the dry 1954–56 period with an extended ENSO(−)–PDO(−) period suggests a link between that particular ocean state and severe drought. Early 1957 saw an abrupt shift to ENSO(+)–PDO(+) conditions over the Pacific. ENSO(+) conditions between April 1957 and June 1958 mark the 1957/58 El Niño event, while the period between June 1957 and June 1960 was dominated by PDO(+) conditions. Accompanying this transition was a similarly abrupt shift in U.S. rainfall. While three of the eight driest years of 1932–99, as determined by November–October NPCP rankings, occurred during 1954, 1955, and 1956, November–October 1957 and 1958 were the 17th and 14th wettest years, respectively. Thus the drought of the 1950s would appear to share at least one circumstantial trait with the 1930s drought, in that both appeared to have terminated with major El Niño events occurring in conjunction with PDO(+) conditions.

c. Cool period of 1964–79

Relatively cool annual temperatures over the United States between 1964 and 1979 produced the most spatially extensive climate impact found in either Figs. 3b,d or 4b,d. The MWZ mapping for temperature during that period can be found in Fig. 8, which shows a significant incidence of low ranked temperatures over most climate divisions outside of the western United States. The year in which the greatest percentage of U.S. area (18%) experienced the coldest year of 1932–99 was 1979. Other years in which relatively large fractions of the continental United States experienced cold annual temperature rankings included 1964, 1965, 1975, and 1978.

The NHT values in Fig. 5d suggest that the 1964–79 cool period over the United States was part of a broader Northern Hemisphere temperature response. Of the 16 yr during 1964–79, the majority of NHT anomalies were negative and six were NHT(−), with the first of those hemispherically cool years occurring in 1964. During the 1970s the prospect of global cooling was addressed in popular science literature and reported in the press (Ponte 1976; New York Times 16 February 1975). A National Academy of Sciences report (National Academy of Sciences 1975) suggested the possibility that the current Holocene interglacial period was nearing its end, but also acknowledged the possible warming effects of increasing atmospheric CO2 levels.

d. Late century wet and warm periods

1) Precipitation

In Fig. 3b the relatively low peak area percentages resulting from the 24- and 25-yr analyses suggests a break between the 26–30-yr analyses and that of the 16–23-yr analyses. This in turn points to the possibility of two response periods in late-century U.S. rainfall. The periods indicated by the 26–30-yr analyses all end in 1998 or 1999, but begin between 1969 and 1973, which indicates the initial onset of wet conditions during the early 1970s. The periods indicated by the 16–23-yr analyses also end in either 1998 or 1999, but begin between 1977 and 1983, which suggests the onset of a second, and possibly distinct wet period in the late 1970s or early 1980s. Figure 3a shows 1972–98 as the peak wet period in the 27-yr analysis, and the MWZ mapping for that period's annual rainfall was introduced before in Fig. 2b. The MWZ mappings for peak wet periods marked by the 26-, 28-, 29-, and 30-yr analyses (not shown) are closely similar to Fig. 2b. Figures 3a and 3b also show the 18-yr period 1982–99 as a peak wet period in the 16–23-yr precipitation analyses, and the MWZ mapping for that time window can be found in Fig. 9a. In Fig. 2b regions that saw a significant incidence of high ranked annual rainfall during 1972–98 are concentrated in the Southwest and along a band of climate divisions extending between the Gulf Coast and the Northeast. In Fig. 9a the 1982–99 MWZ statistics for climate divisions in that eastern band appear to have fallen below a 90% confidence level, while rankings for climate divisions in the Midwest and Great Plains have risen into significance.

In considering the possible roots of late-century wetness over the continental United States, attention is naturally drawn to the shift to mainly positive PDO conditions after 1976 evident in Fig. 5b and noted by many others (e.g., Trenberth 1990; Mantua et al. 1997; Zhang et al. 1997). The configuration of PDO(+) sea surface temperature anomalies is consistent with nationally wet conditions, as cold SST anomalies in the PDO1 region have the potential of drawing the northern Pacific storm track south, and over warmer waters in the vicinity of the PDO2 area and along the west coast of the United States (Namias et al. 1988; Trenberth and Hurrell 1994). But in addition to a shift to more frequent PDO(+) conditions, the years after 1976 also saw high amplitude or persistent ENSO(+) periods occurring during 1982/83, 1986–early 1988, 1991/92, and 1997/98. The increased frequency of ENSO(+) and PDO(+) conditions after 1976 provide likely suspects as sources of increased wetness, particularly in view of the enhanced impact of warm-phase ENSO climate effects during PDO(+) conditions noted by Gershunov and Barnett (1998) and the circumstantial connection between those SSTA conditions and wet years found here (i.e., 1941, 1957, and 1958). But ENSO(+) and PDO(+) conditions may not be the only potential source of wetness over the United States since the early 1970s, as the beginning of that wet regime, that is, the 1969–73 time window indicated by the 26–30-yr analyses, appears to have preceded the PDO shift of the mid-1970s. Moreover, a dry to wet transition over the eastern United States between the 1960s and 1970s suggests the possibility of Atlantic influences.

The relatively regional impact of drought conditions in the northeastern United States during the 1960s was unresolved in the national analysis that led to Figs. 3 and 4. However, when a similar analysis is conducted over climate divisions limited to the eastern third of the United States, that dry period and a transition to wet conditions during the early 1970s become apparent. The counterpart to Fig. 3 resulting from that study (not shown) shows 1963–69 as a period of dry years over the eastern United States, and 1972–79 as a wet period. The MWZ mappings for those time windows can be found in Figs. 10a and 10b, and show a shift from significantly low ranked to high ranked annual rainfall over northeastern climate divisions. The 1960s drought is addressed in some depth by Namias (1966), who attributed the primary cause of that dry period to cool SSTA conditions off the Atlantic seaboard and an associated eastward shift of the storm track into the western Atlantic. Barlow et al. (2001) propose that drought over the eastern United States during the 1960s was the result of a downstream atmospheric response to Northern Pacific sea surface temperature anomalies. But 1963–69 was also dominated by the negative phase of the North Atlantic Oscillation, as five of the seven years between 1963 and 1969 were NAO(−) (Fig. 10c). After 1969, winter NAO values reversed phase during the early 1970s, and wet conditions returned to the northeast in 1972. The NAO can influence climate over the eastern United States, as NAO(−) conditions produce a more zonal storm track over the western North Atlantic, while NAO(+) conditions are associated with a southwesterly storm track more capable of advecting moisture from subtropical regions (Rogers 1990; Hurrell 1995). However, while there is an interesting association in Fig. 10 between NAO(−) conditions and the 1960s drought period, similarly coherent NAO(+) conditions are not evident during 1972–79. As a result, a circumstantial connection between the state of winter NAO and increased annual precipitation in the east during that time seems harder to support.

Although the cause of the shift from a dry 1960s to a wet 1970s over the northeastern United States may be unclear, the expression that it produced in U.S. precipitation indicated in Fig. 10 is quite clear. In addition, those figures, which map that period's MWZ statistics for the entire country, show that significant effects were limited to the eastern third of the United States. The fact that highly ranked annual rainfall during 1972–79 is confined to the east in Fig. 10b might explain the difference in eastern significance in Figs. 2b and 9a. This explanation is based on the idea, as proposed earlier, that the 1972–98 wet regime mapped by Fig. 2b may have occurred in two somewhat distinct stages: a wet period in the east during 1972–79, followed by significantly wet conditions in the central and southwestern United States after the early 1980s. Figure 10b illustrates eastern wetness during 1972–79, while Fig. 9a shows wetness in the southwest and central United States during 1982–99. Although a significant incidence of wet years is not evident in the east during 1982–99, the addition of the wet years of the 1970s raises eastern climate divisions into significance in the 1972–98 analysis of Fig. 2b, which suggests that eastern rainfall rankings during 1982–99 may have been only marginally insignificant (i.e., MWZ ∼ <1.645).

2) Temperature

In Fig. 4b 1986–99 and 1986–92 are peak warm periods in the 14- and 7-yr temperature analyses. The MWZ mappings for temperature rankings during those periods can be found in Figs. 9b and 9c. In both of those figures evidence of warmth is found in the western and northwestern United States, with 1986–92 showing additional evidence of warming over the northern Great Plains. Thus these results generally confirm the western warming implied in the trend analyses of Lettenmaier et al. (1994) and Karl et al. (1996). As measured by the percentage of U.S. area showing annual temperature rankings in the 52–68 range, 1998 and 1999 were the third and second warmest years of 1932–99. Like the cool temperature regime of 1964–79, the warming over the western United States after the mid-1980s occurred roughly concurrently with a similar temperature transition over the Northern Hemisphere, evident in Fig. 5d as a continuous sequence of NHT(+) values between 1987 and 1998.

However, while Lettenmaier et al. (1994) and Karl et al. (1996) also reported cooling trends over the southeastern United States, no significant incidence of cold years are evident over those climate divisions in Fig. 9b during 1986–99. Assuming a semimonotonic decrease in annual temperature over these regions between 1932 and 1999, low-ranked annual temperatures might be expected at the end of the century. The lack of such a result suggests something other than an idealized linear decrease. Inspection of annual temperature series from southeastern climate divisions reveals more of a stepfunction decrease, with warm temperatures during the 1930s (Fig. 6b) and 1950s (Fig. 7b) giving way to generally cooler temperatures after the late 1950s. But when annual temperature trends are calculated for the period 1957–99 (Fig. 11) neutral and positive trends—but no negative trends—are evident in the southeast. Thus negative temperature trends reported elsewhere over this region may be more a result of warmth in the 1930s and 1950s than of a continuous tendency to long-term cooling.

5. Summary and discussion

Annual average temperature and precipitation totals derived from U.S. Climate Division Data were subjected here to statistical analysis based on an exhaustive evaluation of Mann–Whitney U statistics. The primary goal was to objectively survey the timing and location of intra- to multidecadal (IMD) climate impacts over the continental United States during 1932–99. More circumstantial consideration was given to the state of ENSO, the Pacific decadal oscillation, the winter state of the North Atlantic Oscillation, and mean annual Northern Hemisphere surface temperature over that 68-yr period (Figs. 5, 10c). The overall goal was to provide context and insight to both historic and more recent IMD climate variation. The results reported on here include the following.

  • A highly significant tendency to wetter conditions nationally since the early 1970s (Figs. 1, 2 and Table 1), with wetness confined mainly to the east during 1972–79 (Fig. 10b) and a significant incidence of highly ranked water years in climate divisions in the southwest and central United States during 1982–99 (Fig. 9a).
  • After the mid-1980s a significant incidence of highly ranked annual temperatures is evident in the West and the northern Great Plains (Figs. 9b,c), but no evidence of cooling trends were found in the Southeast during the latter half of the twentieth century (Fig. 11), as suggested elsewhere. This late warm period, and a period of cool conditions during 1964–79 that produced the most extensive climate impact found here (Fig. 8), were approximately coincident with temperature anomalies of similar sign averaged over Northern Hemisphere surface areas.
  • Anecdotal evidence suggests that wet years associated with ENSO warm phase conditions and the positive phase of the Pacific decadal oscillation (i.e., cool north central Pacific, warm coastal SST in the northeastern Pacific) may have been influential in breaking the prolonged drought periods of the 1930s and 1950s.
  • Conversely, the La Niña–like signature found here in annual rainfall and temperature during the late 1940s to mid-1950s (Figs. 7a,b), and the increased incidence of cold-phase ENSO and negative-phase PDO conditions during that time (Figs. 5b,c), suggests that ocean–atmosphere conditions associated with that Pacific SST state may play a role in triggering or sustaining periods of severe drought.

A recurring theme in U.S. climate between the 1930s and the late 1960s, and a climate impact of critical interest to society and agriculture, has been prolonged drought. Severe drought also figures prominently in projections of future climate conditions over agriculturally important areas of the central United States (Manabe and Wetherald 1987; Rind et al. 1990; Rosenzweig and Hillel 1993; Gregory et al. 1997). The fact that this region has experienced a significant incidence of highly ranked annual rainfall during 1982–99 (Fig. 9a) suggests a reduced vulnerability to prolonged drought. In addition, that period has also been marked by a persistent shift to Pacific SSTA conditions that have been more consistent with drought breaking than drought making in the past, specifically, the ENSO(+)–PDO(+) conditions that were concurrent with the nationally wet years of 1941, 1957, and 1958. However, conclusions regarding the current drought vulnerability of U.S. climate are difficult to make here given the nature and time resolution of this analysis. Although some areas of the Great Plains and Midwest have shown signs of wetness, periods of significant warmth are also evident over the northern Great Plains (Fig. 9c). Independent evidence of warmer and wetter annual conditions cannot support conclusions regarding long-term drought tendencies, as drought conditions are a function of the simultaneous balance between evaporation and precipitation. Also, the seasonality of those balances are important, as anomalous winter and summer warming leads to different evaporative rates.

The results found here and in previous work (Namias 1982, 1989, 1991; Trenberth and Branstator 1992) supports the idea that drought may have diverse roots, with interrelated local and remote causes. However, the circumstances of the three drought periods discussed here also suggest the potential for either forestalling or forecasting long-term meteorological drought. It seems possible that the persistence of drought conditions over the Great Plains during the 1930s may have been linked to widespread changes in land-use practices during the previous decade. The suggestion made here of large-scale soil moisture deficits over the western Great Plains is speculative, and how those deficits might lead to subsequent long-term rainfall deficits remains an active research topic. Even so, Figs. 6a and 6c provide interesting—if circumstantial—evidence of cause and effect. To forestall such effects, agricultural policy and disincentives could be designed to discourage agricultural expansion into marginal rainfall regions where irrigation is not an option. The other two periods—the droughts of the 1950s and the 1960s—appear to be associated with low-frequency variation in the coupled ocean–atmosphere system. If the physics of those intra- to multidecadal mechanisms were well understood and correctly incorporated into coupled ocean–atmosphere models, it seems possible that the remote components of drought genesis and persistence could be forecast. Given the influence implied here of joint ENSO–PDO variation on U.S. drought, the ability to qualitatively reproduce that variation might be suggested as an important performance benchmark for those coupled models.

Acknowledgments

Thanks to Melissa Nance for help with some of the illustrations. All figures were produced using Generic Mapping Tools (Wessel and Smith 1995).

REFERENCES

  • Akaike, H., 1974: A new look at the statistical model identification. IEEE Trans. Autom. Control, AC-19. 716723.

  • Barlow, M., , S. Nigam, , and E. H. Berbery, 2001: ENSO, Pacific decadal variability, and U.S. summertime precipitation, drought, and streamflow. J. Climate, 14 , 21052128.

    • Search Google Scholar
    • Export Citation
  • Charney, J. G., , W. J. Quirk, , S. Chow, , and J. Kornfield, 1977: A comparative study of the effects of albedo change on drought in semi-arid regions. J. Atmos. Sci., 34 , 13661385.

    • Search Google Scholar
    • Export Citation
  • Dirmeyer, P. A., 1994: Vegetation stress as a feedback mechanism in midlatitude drought. J. Climate, 7 , 14631483.

  • Gershunov, A., , and T. P. Barnett, 1998: Interdecadal modulation of ENSO teleconnections. Bull. Amer. Meteor. Soc., 79 , 27152725.

  • Gershunov, A., , T. Barnett, , and D. Cayan, 1999: North Pacific interdecadal oscillation seen as factor in ENSO-related North American climate anomalies. Eos, Trans. Amer. Geophys. Union, 80 , 25.

    • Search Google Scholar
    • Export Citation
  • Glantz, M. H., 1994: Drought Follows the Plow. Cambridge University Press, 175 pp.

  • Gregory, J. M., , J. F. B. Mitchell, , and A. J. Brady, 1997: Summer drought in northern midlatitudes in a time-dependent CO2 climate experiment. J. Climate, 10 , 662686.

    • Search Google Scholar
    • Export Citation
  • Groisman, P. Y., , and D. R. Easterling, 1994: Variability and trends of total precipitation and snowfall over the United States and Canada. J. Climate, 7 , 184205.

    • Search Google Scholar
    • Export Citation
  • Groisman, P. Y., , and D. R. Legates, 1994: The accuracy of United States precipitation data. Bull. Amer. Meteor. Soc., 75 , 215227.

  • Groisman, P. Y., , R. W. Knight, , and T. R. Karl, 2001: Heavy precipitation and high streamflow in the contiguous United States: Trends in the twentieth century. Bull. Amer. Meteor. Soc., 82 , 219246.

    • Search Google Scholar
    • Export Citation
  • Guttman, N. B., , and R. G. Quayle, 1996: A historical perspective of U.S. climate divisions. Bull. Amer. Meteor. Soc., 77 , 293303.

  • Halpert, M. S., , and C. F. Ropelewski, 1992: Surface temperature patterns associated with the Southern Oscillation. J. Climate, 5 , 577593.

    • Search Google Scholar
    • Export Citation
  • Hollander, M., , and D. A. Wolfe, 1999: Nonparametric Statistical Methods. 2d ed. Wiley and Sons, 787 pp.

  • Hu, Q., , C. M. Woodruff, , and S. E. Mudrick, 1998: Interdecadal variations of annual precipitation in the central United States. Bull. Amer. Meteor. Soc., 79 , 221229.

    • Search Google Scholar
    • Export Citation
  • Hurrell, J. W., 1995: Decadal trends in the North Atlantic Oscillation and relationships to regional temperature and precipitation. Science, 269 , 676679.

    • Search Google Scholar
    • Export Citation
  • Jones, P. D., cited. 2001: Gridded hemispheric and global temperature averages. [Available online at http://www.cru.uea.ac.uk/cru/data/temperature.].

    • Search Google Scholar
    • Export Citation
  • Kaplan, A., , M. Cane, , Y. Kushnir, , A. Clement, , M. Blumenthal, , and B. Rajagopalan, 1998: Analyses of global sea surface temperature 1856–1991. J. Geophys. Res, 103 , 1856718589.

    • Search Google Scholar
    • Export Citation
  • Karl, T. R., , and P. D. Jones, 1989: Urban bias in area-averaged surface temperature trends. Bull. Amer. Meteor. Soc., 70 , 265270.

  • Karl, T. R., , and R. W. Knight, 1998: Secular trends of precipitation amount, frequency, and intensity in the USA. Bull. Amer. Meteor. Soc., 79 , 231241.

    • Search Google Scholar
    • Export Citation
  • Karl, T. R., , L. Metcalf, , M. L. Nicodemus, , and R. Quayle, 1983: Statewide average climatic history. Historical Climatology Series 6-1, National Climate Data Center, 35 pp.

    • Search Google Scholar
    • Export Citation
  • Karl, T. R., , C. N. Williams Jr., , P. J. Young, , and W. Wendland, 1986: A model to estimate the time of observation bias associated with monthly mean maximum, minimum and mean temperatures for the United States. J. Climate Appl. Meteor., 25 , 145160.

    • Search Google Scholar
    • Export Citation
  • Karl, T. R., and Coauthors. 1993: A new perspective on recent global warming: Asymmetric trends in daily maximum and minimum temperature. Bull. Amer. Meteor. Soc., 74 , 10071023.

    • Search Google Scholar
    • Export Citation
  • Karl, T. R., , R. W. Knight, , and N. Plummer, 1995: Trends in high frequency climate variability during the 20th century. Nature, 377 , 217220.

    • Search Google Scholar
    • Export Citation
  • Karl, T. R., , R. W. Knight, , D. R. Easterling, , and R. G. Quayle, 1996: Indices of climate change for the United States. Bull. Amer. Meteor. Soc., 77 , 279292.

    • Search Google Scholar
    • Export Citation
  • Kiladis, G. N., , and H. F. Diaz, 1989: Global climatic anomalies associated with extremes in the Southern Oscillation. J. Climate, 2 , 10691090.

    • Search Google Scholar
    • Export Citation
  • Kunkel, K. E., , K. Andsager, , and D. R. Easterling, 1999: Long-term trends in extreme precipitation events over the conterminous United States and Canada. J. Climate, 12 , 25152527.

    • Search Google Scholar
    • Export Citation
  • Langbein, W. B., , and J. R. Slack, 1982: Yearly variations in runoff and frequency of dry years for the conterminous United States. U.S. Geological Survey Open File Rep. 82-751, 85 pp.

    • Search Google Scholar
    • Export Citation
  • Lau, K. M., , and H. Weng, 1995: Climate signal detection using wavelet transform: How to make a time series sing. Bull. Amer. Meteor. Soc., 76 , 23912402.

    • Search Google Scholar
    • Export Citation
  • Legates, D. R., 1995: Precipitation measurement biases and climate change detection. Preprints, Sixth Symp. on Global Change Studies, Dallas, TX, Amer. Meteor. Soc., 168–173.

    • Search Google Scholar
    • Export Citation
  • Lettenmaier, D. P., , E. F. Wood, , and J. R. Wallis, 1994: Hydro-climatological trends in the continental United States, 1948–88. J. Climate, 7 , 586607.

    • Search Google Scholar
    • Export Citation
  • Lins, H. F., , and J. R. Slack, 1999: Streamflow trends in the United States. Geophys. Res. Lett., 26 , 227230.

  • Livezey, R. E., , M. Masutani, , A. Leetma, , H. Rui, , M. Ji, , and A. Kumar, 1997: Teleconnective response of the Pacific–North American region to large central equatorial Pacific SST anomalies. J. Climate, 10 , 17871820.

    • Search Google Scholar
    • Export Citation
  • Lockeretz, W., 1978: The lessons of the dust bowl. Amer. Sci., 66 , 560569.

  • Manabe, S., , and R. T. Wetherald, 1987: Large-scale changes of soil wetness induced by an increase in atmospheric carbon dioxide. J. Atmos. Sci., 44 , 12111235.

    • Search Google Scholar
    • Export Citation
  • Mann, H. B., , and D. R. Whitney, 1947: On a test of whether one of two random variables is stochastically larger than the other. Ann. Math. Stat., 18 , 5060.

    • Search Google Scholar
    • Export Citation
  • Mantua, N. J., , S. R. Hare, , Y. Zhang, , J. M. Wallace, , and R. C. Francis, 1997: A Pacific interdecadal climate oscillation with impacts on salmon production. Bull. Amer. Meteor. Soc., 78 , 10691079.

    • Search Google Scholar
    • Export Citation
  • Mauget, S. A., , and D. R. Upchurch, 1999: El Niño and La Niña related climate and agricultural impacts over the Great Plains and Midwest. J. Prod. Agric., 12 , 203214.

    • Search Google Scholar
    • Export Citation
  • Mendenhall, W., , D. D. Wackerly, , and R. L. Sheaffer, 1990: Mathematical Statistics with Applications. 4th ed. PWS-Kent, 788 pp.

  • Mote, T., 1996: Influence of ENSO on maximum, minimum, and mean temperatures in the southeast United States. Phys. Geogr., 17 , 497512.

    • Search Google Scholar
    • Export Citation
  • Namias, J., 1966: Nature and possible causes of the northeastern United States drought during 1962–1965. Mon. Wea. Rev., 94 , 543554.

    • Search Google Scholar
    • Export Citation
  • Namias, J., 1982: Anatomy of Great Plains protracted heat waves (especially the 1980 U.S. summer drought). Mon. Wea. Rev., 110 , 824838.

    • Search Google Scholar
    • Export Citation
  • Namias, J., 1989: Cold waters and hot summers. Nature, 338 , 1516.

  • Namias, J., 1991: Spring and summer drought over the contiguous U.S.—Causes and prediction. J. Climate, 4 , 5465.

  • Namias, J., , X. Yuan, , and D. R. Cayan, 1988: Persistence of North Pacific sea surface temperature and atmospheric flow patterns. J. Climate, 1 , 682703.

    • Search Google Scholar
    • Export Citation
  • National Academy of Sciences, 1975: Understanding Climatic Change: A Program for Action. National Academy of Sciences, 239 pp.

  • National Research Council, 2002: Abrupt Climate Change: Inevitable Surprises. National Academy Press, 244 pp.

  • Oglesby, R. J., , and D. J. Erickson, 1989: Soil moisture and the persistence of North American drought. J. Climate, 2 , 13621380.

  • Palmer, T. N., , and C. Brankoviĉ, 1989: The 1988 U.S. drought linked to anomalous sea surface temperature. Nature, 338 , 5457.

  • Parker, D. E., , P. D. Jones, , C. K. Folland, , and A. Bevan, 1994: Interdecadal changes of surface temperature since the late 19th century. J. Geophys. Res., 99 , 1437314399.

    • Search Google Scholar
    • Export Citation
  • Ponte, L., 1976: The Cooling. Prentice-Hall, 268 pp.

  • Puckett, W. E., , J. H. Dane, , and B. F. Hajek, 1985: Physical and mineralogical data to determine soil hydraulic properties. Soil Sci. Soc. Amer. J., 49 , 831836.

    • Search Google Scholar
    • Export Citation
  • Quayle, R. G., , D. R. Easterling, , T. R. Karl, , and P. J. Hughes, 1991: Effects of recent thermometer changes in the cooperative station network. Bull. Amer. Meteor. Soc., 72 , 17181723.

    • Search Google Scholar
    • Export Citation
  • Rind, D., 1982: The influence of ground moisture conditions in North America on summer climate as modeled in the GISS GCM. Mon. Wea. Rev., 110 , 14871493.

    • Search Google Scholar
    • Export Citation
  • Rind, D., , R. Goldberg, , J. Hansen, , C. Rosenzweig, , and R. Ruedy, 1990: Potential evaporation and the likelihood of future drought. J. Geophys. Res., 95 , (D7),. 998310004.

    • Search Google Scholar
    • Export Citation
  • Rogers, J., 1990: Patterns of low-frequency monthly sea level pressure variability (1899–1986) and associated wave cyclone frequencies. J. Climate, 3 , 13641379.

    • Search Google Scholar
    • Export Citation
  • Ropelewski, C. F., , and M. S. Halpert, 1989: Precipitation patterns associated with the high index phase of the Southern Oscillation. J. Climate, 2 , 268283.

    • Search Google Scholar
    • Export Citation
  • Rosenzweig, C., , and D. Hillel, 1993: The Dust Bowl of the 1930's: Analog of greenhouse effect in the Great Plains? J. Environ. Qual., 22 , 922.

    • Search Google Scholar
    • Export Citation
  • Schaal, L. A., , and R. F. Dale, 1977: Time of observation temperature bias and “climatic change.”. J. Appl. Meteor., 16 , 215222.

  • Shabbar, A., , and M. Khandekar, 1996: The impact of El Niño–Southern Oscillation on the temperature field over Canada. Atmos.–Ocean, 34 , 401416.

    • Search Google Scholar
    • Export Citation
  • Shukla, J., , and Y. Mintz, 1982: Influence of land-surface evapotranspiration on the Earth's climate. Science, 215 , 14981500.

  • Stephenson, D. B., cited. 1999: The North Atlantic Oscillation thematic website. [Available online at http://www.met.rdg.ac.uk/cag/NAO/.].

    • Search Google Scholar
    • Export Citation
  • Tannehill, I. R., 1947: Drought, Its Causes and Effects. Princeton University Press, 264 pp.

  • Thiebaux, H. J., , and F. W. Zwiers, 1984: The interpretation and estimation of effective sample size. J. Climate Appl. Meteor., 23 , 800811.

    • Search Google Scholar
    • Export Citation
  • Trenberth, K. E., 1990: Recent observed climate changes in the Northern Hemisphere. Bull. Amer. Meteor. Soc., 71 , 988993.

  • Trenberth, K. E., , and G. W. Branstator, 1992: Issues in establishing causes of the 1988 drought over North America. J. Climate, 5 , 159172.

    • Search Google Scholar
    • Export Citation
  • Trenberth, K. E., , and J. W. Hurrell, 1994: Decadal atmosphere–ocean variations in the Pacific. Climate Dyn., 9 , 303319.

  • Trenberth, K. E., , G. W. Branstator, , and P. A. Arkin, 1988: Origins of the 1988 North American Drought. Science, 242 , 16401645.

  • Wessel, P., , and W. H. F. Smith, 1995: New version of the Generic Mapping Tools released. Eos, Trans. Amer. Geophys. Union, 76 , 329.

  • Wilcoxon, F., 1945: Individual comparisons by ranking methods. Biometrics Bull., 1 , 8083.

  • Wilks, D. S., 1995: Statistical Methods in the Atmospheric Sciences. Academic Press, 464 pp.

  • Wolfson, N., , R. Atlas, , and Y. C. Sud, 1987: Numerical experiments related to the summer 1980 heat wave. Mon. Wea. Rev., 115 , 13451357.

    • Search Google Scholar
    • Export Citation
  • Woodhouse, C. A., , and J. T. Overpeck, 1998: 2000 years of drought variability over the central United States. Bull. Amer. Meteor. Soc., 79 , 11451162.

    • Search Google Scholar
    • Export Citation
  • Worster, D., 1979: The Dust Bowl: The Southern Plains in the 1930's. Oxford University Press, 277 pp.

  • Zhang, Y., , J. M. Wallace, , and D. S. Battisti, 1997: ENSO-like interdecadal variability: 1900–93. J. Climate, 10 , 10041020.

Fig. 1.
Fig. 1.

(a) Time series of cumulative Nov–Oct rainfall spatially averaged over the 343 climate divisions of the coterminous United States (NPCP) for 1896–1999. Black bars indicate the 10 most highly ranked water years. Water years are identified with calendar year containing the final ten months of each Nov–Oct period; i.e., 1896 refers to Nov 1895 to Oct 1896, etc. (b) Mann–Whitney Z statistics for running 27-yr samples of NPCP rankings. Horizontal lines indicate positive and negative significance at 90%, 95%, and 99% confidence levels

Citation: Journal of Climate 16, 13; 10.1175/2751.1

Fig. 2.
Fig. 2.

(a) Percentage of U.S. area showing significant (>1.645) Mann–Whitney Z (MWZ) statistics for Nov–Oct rainfall over running 27-yr time windows during 1932–99, with rankings calculated over the 1932–99 base period. (b) MWZ statistics for annual rainfall rankings at the climate division level during 1972–98. Gray shaded and hatched climate divisions show statistics significant at 90%, 95%, and 99% confidence levels

Citation: Journal of Climate 16, 13; 10.1175/2751.1

Fig. 3.
Fig. 3.

(a) N-year time windows during which relatively large areas of the United States experienced a significant incidence of high (MWZ > 1.645) rankings in annual rainfall, as determined by timelines similar to Fig. 2a. Black bars show peak wet periods at each sample size for N = 6 to 30 yr. Gray bars mark secondary peak periods, defined here as periods during which significant Z statistics were found over areas comparable to peak area values, but not overlapping the primary peak period in time. (b) Percentage of total U.S. area showing significance at a 90% or better confidence level during the corresponding peak wet periods in (a). (c) As in (a) for peak and secondary periods marked by a significant incidence of low (i.e., MWZ < −1.645) rankings in annual rainfall. (d) Percentage of U.S. area showing significant MWZ statistics during corresponding peak dry periods in (c)

Citation: Journal of Climate 16, 13; 10.1175/2751.1

Fig. 4.
Fig. 4.

(a) As in Fig. 3a for N-year time windows marked by a significant incidence of high rankings in annual temperature. (b) Percentage of total U.S. area showing significance at a 90% or better confidence level during the corresponding peak warm periods in (a). (c) As in (a) for peak and secondary periods marked by a significant incidence of low rankings in annual temperature. (d) Percentage of U.S. area showing significant MWZ statistics during corresponding peak cool periods in (c)

Citation: Journal of Climate 16, 13; 10.1175/2751.1

Fig. 5.
Fig. 5.

(a) Timelines of significant dry, wet, cool, and warm periods as indicated by Figs. 3 and 4. (b) 1920–99 monthly cold tongue SSTA anomalies as derived from Kaplan et al. (1998) analysis values and regression analysis. (c) Composite PDO index formed from unfiltered and normalized SSTA analysis values averaged over the PDO1 and PDO2 regions outlined in (e). (d) 1920–99 annual mean Northern Hemisphere surface temperature anomaly (Parker et al. 1994). (e) First unrotated EOF of low-frequency (υ < 72 months−1) northern Pacific SSTA analyses over 1920–91, and cold tongue region as defined here

Citation: Journal of Climate 16, 13; 10.1175/2751.1

Fig. 6.
Fig. 6.

(a) As in Fig. 2b for annual rainfall rankings during 1933–40. (b) As in Fig. 2b for annual temperature rankings during 1932–39. (c) Increase in acreage of harvested crops between 1919 and 1929 (from Tannehill 1947)

Citation: Journal of Climate 16, 13; 10.1175/2751.1

Fig. 7.
Fig. 7.

(a) As in Fig. 2b for annual rainfall rankings during 1950–56. (b) As in Fig. 2b for annual temperature rankings during 1949–56

Citation: Journal of Climate 16, 13; 10.1175/2751.1

Fig. 8.
Fig. 8.

As in Fig. 2b for annual temperature rankings during 1964–79

Citation: Journal of Climate 16, 13; 10.1175/2751.1

Fig. 9.
Fig. 9.

(a) As in Fig. 2b for annual rainfall rankings during 1982–99. (b) As in Fig. 2b for annual temperature rankings during 1986–99. (c) As in Fig. 2b for annual temperature rankings during 1986–92

Citation: Journal of Climate 16, 13; 10.1175/2751.1

Fig. 10.
Fig. 10.

(a) As in Fig. 2b for annual rainfall rankings during 1963–69. (b) As in Fig. 2b for annual rainfall rankings during 1972–79. (c) Northern winter (DJFM) North Atlantic Oscillation values for 1920–98 (Hurrell 1995).

Citation: Journal of Climate 16, 13; 10.1175/2751.1

Fig. 11.
Fig. 11.

Trend variation in annual temperatures during 1957–99. Temperature values indicated are trend coefficients derived from linear trend analysis of temperature during 1957–99, multiplied times 43 (yr)

Citation: Journal of Climate 16, 13; 10.1175/2751.1

Table 1.

Results for hypergeometric and Mann–Whitney Z analyses of rankings of nationally averaged annual precipitation values (NPCP) for both 1896–1999 and 1932–99 base periods, and for NPCP values accumulated over both Nov–Oct and Sep–Oct water years. Peak Z periods refer to the 27-yr period resulting in the highest Z statistics in running analyses similar to Fig. 1b

Table 1.

1

Those years, and their NPCP rankings, are 1973(1), 1975(10), 1979(6), 1983(3), 1984(8), 1993(2), 1995(9), 1997(7), and 1998(5).

2

Water years are identified with calendar year containing the final ten months of each November–October period; i.e., 1896 refers to November 1895 to October 1896, etc.

3

Winter NAO years refer to the year of the January of the DJFM season.

Save