The Southeast Drought in Historical Perspective Abstract The dry and hot weather in the southeast United States during the first seven months of caused record drought. The agricultural and hydrological perspectives of this drought are examined via a climatological time series. Late nineteenth and twentieth century climate data from the most severely affected areas indicate that from an agricultural perspective the beginning and middle of the growing season was by far the worst on record. On the other hand, from a hydrological perspective the drought was not of sufficient duration to stand out as such an extreme anomaly. The drought is part of a change in recent years from the wet weather of the 1960s and much of the 1970s. At this time, there is no evidence to suggest that this change is anything more than another in a series of climate fluctuations typical throughout the climate records of many areas. 1. Introduction In order to facilitate the subsequent assessment of the drought it is important that we understand what is meant by the word "drought." The word itself often has different meanings to various groups depending upon their specific interests. We will refer to two aspects of drought. The first aspect will pertain to the hydrological drought or the long-term water deficiency in deep-soil profiles. This aspect of drought is important with respect to regional, municipal, and localwater supplies used for domestic, commercial, and industrial processes. The hydrological drought usually requires at least several months of dry weather to develop. Likewise, it normally takes an extended period of wetness (months to years) before the hydrological drought ends, returning water levels in streams, lakes, and ground water to near normal. The second aspect of drought that will be addressed pertains to a short-term moisture deficiency in shallow plantroot zones or the so-called agricultural drought. By definition, this occurs whenever the vegetation of an area is stressed due to an inadequate or below-normal supply of moisture. This type of drought can develop rather suddenly compared to the hydrological drought. Sometimes just a few weeks without adequate moisture in the Southeast, depending on the timing during the growing season, can often lead to reduced crop yields and stress on vegetation. It is important to understand that the hydrological drought may persist during several waxings and wanings of the agricultural drought. Likewise, the onset and termination of an agricultural drought can occur in the midst of a hydrological wet spell. Two further clarifications are required. We will discuss the Southeast Drought f r o m a climatological perspective. 1987 American Meteorological Society Bulletin American Meteorological Society Thomas R. Karl and Pamela J. Young National Climatic Data Center Asheville, North Carolina Note that for equal anomalies of dry weather separated in time, the human impact of these events may not necessarily be the same. Changes in water demand, i.e., increased usage for industrial, commercial, or domestic purposes, or changes in water supply and distribution, i.e., new reservoirs, losses in water-supply lines, new water-supply lines, etc., are not accounted for in a climatological perspective. Additionally, it must be recognized that drought is relative to some expected norm. For example, our natural environment and our local economies in the Southeast are based on a relatively ample supply of rainfall compared to drier portions of our nation such as the Western Great Plains or the Southwest. Using this concept of drought, the departure f r o m the expected moisture supply is all important, rather than the absolute amount of precipitation, in order to compare and contrast the temporal characteristics and spatial characteristics of droughts. 2. The initiation and development of the drought During the fall of 1985, largely due to a very wet October and November, much of the southeastern United States was able to pull itself out of the hydrological drought that had developed over the area during the past year. This included the Carolinas (North Carolina and South Carolina); the Virginias (Virginia and West Virginia), Georgia, parts of Alab a m a, and Florida, as well as the eastern half of Tennessee. During the subsequent winter months, December 1985 through February, a lack of Gulf Coast and East Coast winter storms resulted in the driest winter of the twentieth century in NC; the second driest in Mississippi, Louisiana, and Tennessee, and the fifth, sixth, and seventh driest winters in South Carolina, Alabama, and Kentucky, respectively. As a result, by the end of winter a moderate hydrological drought had redeveloped from Louisiana northeast to southern Virginia. Conditions continued to deteriorate during the beginning of the growing season in the Southeast. March through May was the second driest spring of the twentieth century in North Carolina, the third driest in Tennessee and Virginia, the sixth driest in Kentucky, and the eighth driest in South Carolina. These conditions intensified the hydrological drought leading to a simultaneous occurrence of a severe agricultural drought in many of these states. Precipitation continued to be well-below normal across Alabama, South Carolina, North Carolina, Virginia, and Georgia during the first two months of summer as the driest June of the twentieth century was recorded across the state of Virginia, and the seventh driest in North Carolina, South Carolina, and Georgia. Temperatures during these months were much above 773
774 Vol. 68, No. 7, July 1987 normal, increasing the demand for water. The warmest July of the twentieth century was recorded in the states of North Carolina, South Carolina, and Georgia while Tennessee, Virginia, and Alabama had their sixth or seventh warmest July. Precipitation remained well-below normal during July in all of these states with South Carolina and Georgia having their third and fourth driest Julys, respectively. The dry and hot weather produced the most extreme agricultural drought of record in many of these areas. The hydrological drought continued to worsen through July when it reached its maximum areal extent. Streamflow across most of the Southeast was less than 50 percent of normal during these months. Beginning in August and continuing through the fall and winter seasons wetter conditions abated both the agricultural and hydrological droughts in most areas of the Southeast. 3. The hydrological drought in historical perspective In order to make some quantitative inferences regarding the severity of the hydrological drought a model has been developed (Palmer, 1965; Karl, ) that produces an index of hydrologic drought severity (calibrated 1931-85). This index, the Palmer Hydrological Drought Index (PHDI), identifies the month of drought initiation as well as termination. During each drought episode precipitation deficiencies can also be calculated in terms of accumulated precipitation deficits f r o m normal. In order to put the recent drought in perspective, we have used the model to evaluate the climatological data back to the turn of the century. This has been completed for nine of the most severely affected climate divisions of the Southeast: 1) North Central, Georgia; 2) Northeast, Georgia; 3) Central, Georgia; 4) Southern Mountains, North Carolina; 5) Northern Mountains, North Carolina; 6) Central Piedmont, North Carolina; 7) Southern Piedmont, North Carolina; 8) Northwest, South Carolina, and; 9) Eastern, Tennessee. Using an appropriate statistical model it is possible to estimate the unusualness of the recent hydrological drought. Using the actual index of the hydrological drought severity as a measure of its intensity the absolute value of the lowest P H D I in each drought episode was identified. Additionally, the cumulative deficit of precipitation during each drought episode ( P H D I < 0.5) was also used as another measure of long-term drought severity. Figure 1 depicts the time series of these quantities. It should be noted that climate-division data prior to 1931 is derived f r o m regression analysis of statewide averages described by Karl et al., (1983) and Karl and Knight (1985). The importance of including these data is suggested by the drought in the 1920s in many areas of the Southeast (Fig. 1). Similar to the procedures described in Karl and Young () various distributions were tested to find the most appropriate statistical model for the data. The maximum-likelihood estimates of the V and L o g - r statistical distribution models were tested using the Kolmorgorov-Smirnov (KS) goodness-of-fit test. The Log-Normal model was not tested as the T distribution can approximate a normal distribution when the shape parameter is large. The results indicated that the T distribution was consistently the best fit to the lowest values of the P H D I during each episode, and the L o g - r distribution was most appropriate for the cumulative precipitation deficits during each of the drought episodes. For each climate division used in the analysis the KS test indicated that the data were consistent with the T distribution. The hypothesis that the data come f r o m a T distribution could not be rejected, even at the 10 percent level. Using these distributions the results indicate that on average, assuming no climate change, a hydrological drought as severe as the drought can be expected to recur about once in every 40 to 100 years for the areas outside of North Carolina, and in excess of 100 years for the regions tested within North Carolina (Table 1). At least half of the streams in the area set new record-low streamflows by the end of July. These streams had records dating back f r o m 48 to 91 years. On the other hand, using the cumulative deficit of precipitation as a measure of drought severity the recurrence intervals for a hydrological TABLE 1. Recurrence intervals with respect to the lowest Palmer Hydrological Drought Index and the cumulative precipitation deficit during the drought indicated. West Central, Minnesota's 1930s drought is included for comparison. Hydrologic Related Drought Climate Division Year N. Cen. GA* NE GA Cen. GA S. Mtn. NC N. Mtn. NC Cen. Pied. NC S. Pied. NC NW SC E. TN W. Cen. MN 1934 PHDI Recurrence Interval Years Precipitation (Deficit) Recurrence Interval Years -5.33-5.05-5.29-6.12-5.44-6.36-6.36-5.21-5.31-10.32 82 85 43 685 293 150 137 75 93 358-20.08-20.50-23.62-32.79-15.55-16.36-15.53-15.93-25.83-47.83 20 17 30 56 23 24 23 18 33 226 * Georgia: North Central, Northeast, Central; North Carolina: Southern Mountains, Northern Mountains, Central Piedmont, Southern Piedmont; South Carolina: Northwest; Tennessee: Eastern; Minnesota: West Central.
Bulletin American Meteorological Society 775 drought as severe as the drought are between 20 and 60 years. The exact year varies with each region (Table 1). Overemphasis of the exact number of years of the recurrence interval should be avoided. Very long recurrence intervals are particularly sensitive to the form of the distribution chosen. Instead, emphasis should be placed on the general characteristics. In terms of the magnitude of the P H D I itself, the modern climate record does not contain any values as low as observed in the western Carolinas, and such an extreme value is to be considered a rare event based on the probability distribution of such events. On the other hand, because the duration of the drought had been relatively short, despite its extreme intensity, the precipitation deficit is not uncharacteristically large. Several other droughts have had cumulative precipitation deficits as large or larger than that observed through July of. Even in the Southern Mountains of North Carolina where the rains in the autumn of 1985 did not end the ongoing drought, the 50 year recurrence interval of the cumulative precipitation deficit cannot be considered a rare event. vere as the recent one would not be expected to recur on average more than once in 100 years (Table 2). For the other regions the recurrence interval is even longer. The average recurrence interval for an agricultural drought as severe as the recent drought in Eastern Tennessee, Northwest South Carolina, and the Southern Mountains of North Carolina is about once in every 200 years. F o r the Central and Southern Piedmont and Northern mountains of North Carolina the recurrence interval of the March through July agricultural drought is on the order of once in every several hundred years. The agricultural drought in these areas was the worst since climatological observations became routinely available (circa 1890) and is truly remarkable in its strength. By way of comparison, one of the most severe hydrological droughts in the United States occurred during the "dustbowl" days of the 1930s in West Central, Minnesota. For that particular hydrological drought the recurrence interval is about once in every 200 to 400 years regardless of how it is viewed. We currently have no evidence to indicate that the recent drought is a result of global increases of carbon dioxide and other greenhouse gases. The recurrence interval for the recent hydrological drought is not sufficiently long to make it inconsistent with the twentieth-century climate record. The agricultural drought during the growing season (March through July) was quite remarkable in its intensity for a few selected areas, but single growing-season events, even moderately rare ones, do not necessarily imply a semipermanent change to a new climate regime. The climate must be viewed on larger time scales. In this regard an unusual characteristic of the precipitation climatology in the Southeast during recent years has been the anomalously wet and cool years of the late 1950s and the decades of the 1960s and 1970s. This wet period coincided to some extent with the greatest change in the United States average precipitation since the dust-bowl days of the 1930s. Since 1970, spring and a u t u m n precipitation across the nation as a whole has increased by about 6 percent and 12 percent respectively, compared to the twentieth-century average. This has lead to nearly a 5 percent increase in annual precipitation. By comparison, on an annual basis the decade of the 1930s in the United States had almost 6 percent less precipitation than the rest of the twentieth century. Yet, areas 4. The agricultural drought in historical perspective In order to estimate the severity of the climate conditions with respect to the agricultural drought in the Southeast another index is used, the moisture-anomaly index. This index is a measure of the short-term moisture deficiency. It is evaluated on a monthly basis and has no memory of past conditions except for the initial amount of moisture in the soil at the start of a month. For each month of the year the index integrates several climate-dependent variables. This includes the expected initial a m o u n t of moisture stored in the soil at the start of a m o n t h, the expected a m o u n t of rainfall runoff and recharge to the soil given the initial condition, and the expected a m o u n t of moisture to be given up by the soil due to evaporation and plant transpiration. The moistureanomaly index, or the Z index as it is commonly called, has been calculated for the United States back to the turn of the century (Karl, ). The mean value of this index through the beginning and middle portions of the growing season (March through July) was calculated for the nine regions most severely affected by the hydrological drought (Fig. 2). By comparing the magnitude of the most recent drought with previous agricultural droughts for the same time period (March through July) it is possible to estimate the recurrence interval of the agricultural drought for each region. G o o d ness-of-fit KS tests indicated that the T distribution was an appropriate model to use to estimate recurrence intervals of the drought. Through the use of the F distribution the model indicates that for the areas in Georgia, an agricultural drought as se- 5. Is the recent drought related to the predicted climate changes from increased carbon dioxide and other greenhouse gases? TABLE 2. Same as Table 1 except for the Z index. Agriculture Related Drought Climate Division N. Cen. GA NE GA Cen. GA S. Mtn. NC N. Mtn. NC Cen. Pied. NC S. Pied. NC NW SC E. TN W. Cen. MN Year Z Index Drought Recurrence Interval 1934-2.80-2.58-3.01-2.92-2.99-3.43-3.60-2.72-2.79-4.72 118 132 114 235 524 477 979 224 235 175
776 Vol. 68, No. 7, July 1987 FIG. 1. Time series of Palmer Hydrological Drought Index and the cumulative precipitation deficit during major drought episodes (PHDI < 3).
Bulletin American Meteorological Society 777 FIG. 2. Time series of the average moisture anomaly (Z index) for the growing season March through July. The two horizontal lines represent the mean and the lowest value on record.
778 Vol. 68, No. 7, July 1987 such as West Central, Minnesota, which were so severely affected by that drought, are now coping with too much water as many lakes in the United States, e.g., the Great Lakes and the Great Salt Lake, are at record high levels. As more and more of the twentieth-century climate records are compiled and processed the evidence seems to indicate that our climate is not only variable from year to year, but often from decade to decade. The explanation of these decadal climate fluctuations is very important if we expect to unequivocally relate ongoing climate anomalies such as the recent drought in the Southeast to increases in greenhouse gases. 6. Summary The agricultural drought in the southeast during the critical March through July period, was the most severe that we have witnessed in the past 90 years, and truly remarkable in its severity. The hydrological drought, which resulted in the lowest observed streamflows in more than half a century, was unusually severe, but not unprecedented in terms of duration. There is no evidence that the drought in the Southeast is a result of increasing carbon dioxide or other trace gases. Rather, it appears that the most recent drought is part of another one of a series of climate fluctuations that are typical of the climate record of the United States throughout the twentieth century. Acknowledgments. This work was partially funded by the National Climate Program office's support of the Climate Impact Perception and Adjustment Experiment (CLIMPAX). References Karl, T. R., : The sensitivity of the Palmer Drought Severity Index (PDSI) and Palmer's Z-index to their calibration coefficients including potential evapotranspiration. J. Clim. Appl. Meteor., 25, 77-86. Karl, T. R., and P. J. Young, : Recent heavy precipitation in the vicinity of the Great Salt Lake: Just how unusual? Bull. Amer. Meteor. Soc., 67, 4-9. Karl, T. R., and R. W. Knight, 1985: Atlas of Monthly and Seasonal Precipitation Departures from Normal (1895-1985) for the Contiguous United States: Spring through Winter. Historical Climatology Series 3-12, 13, 14, 15, National Climatic Data Center, Federal Building, Asheville, NC 28801. Karl, T. R., L. K. Metcalf, M. L. Nicodemus, and R. G. Quayle, 1983: Statewide average climatic history: Alabama through Wyoming. Historical Climatology Series 6-1, National Climatic Data Center, Federal Building, Asheville, NC 28801. Palmer, W. C., 1965: Meteorological drought research paper No. 45. U.S. Weather Bureau, 50 pp. (Available from NOAA Library and Information Services Division, Washington, DC 20852)