2000 years of Drought Variability in the Central United States

Home Browse Search

FIG. 2. Locations of sources of historical drought data for the Great Plains, 1795 -1895. Green shaded areas represent climate regions based on cluster analysis from Mock's (1991) analysis of nineteenth - century climate records. The dates (dark green) represent years in which droughts were reported in more than one region for two or more consecutive seasons. Brown areas are regions of sand dunes and eolian activity, accompanied by the years (in red) in which active sand move- ment was reported (Muhs and Holliday 1995). The gray region represents the gen- eral region of early meteorological stations from which Ludlum (197 1) derived drought years (in blue). Newspaper accounts are from a variety of newspapers in eastern and central Kansas (Bark 1978). eolian activity, and on the time line in Fig. 3. Several periods of eolian activity were reported in many ar- eas between 1840 and 1865, with other intervals in the late 1700s and early 1800s, as well as at the end of the nineteenth century (Muhs and Holliday 1995). Interestingly, although some eolian activity was re- ported in the 1930s and 1950s, these twentieth -cen- tury droughts were not severe or long enough to cause regional mobilization of dunes (Muhs and Maat 1993; Madole 1994; Muhs and Holliday 1995). Numerous reconstructions of precipitation and summer drought have been generated for the Great Plains from tree -ring chronologies located in regions proximal to the Great Plains, shown on the map in Fig. 4a (Table 2 contains the key for the symbols in this figure). Regression - based tree -ring reconstructions of cli- mate tend to underestimate extreme values, a consequence of the regression techniques used in producing the recon- structions. However, dry extremes are better replicated than wet extremes, and reconstructions of drought extent and duration are reasonably accurate. For example, in Fig. 5, a comparison of ob- served and reconstructed mapped Palmer Drought Severity Index (PDSI) (Palmer 1965) values for the severe drought years of 1934 and 1956 shows that drought severity is generally about one PDSI value lower (less severe) for the recon- structed values than for the observed val- ues (Cook et al. 1998). However, the spatial extents of the droughts are well replicated by the reconstructed values, as are drought durations of the 1930s and 1950s events. Although the absolute severity is not duplicated in the tree - ring reconstructions, assessments of the relative severity of twentieth - century droughts compared to droughts in previ- ous centuries can still be made. The amount of variance in the observed drought and precipitation series ex- plained by tree -ring chronologies varies, with average values of about 55 %, rang- ing up to 67% (Table 3). These values are good compared to those obtained in dendroclimatic studies in the semiarid to aria western United States, where trees are notably sensitive to climate. The tree - ring records, of course, are unable to explain all of the drought or precipitation variability because tree growth is usually not solely affected by precipitation or drought conditions (Douglass 1914, 1929). Many of the tree -ring reconstructions suggest that the droughts of the 1930s and 1950s have been equaled or, in some regions, surpassed by droughts in the past several centuries. This is illustrated in the graphs of PDSI reconstructions from Cook et al. (1996) and Cook et al. (1998) for grid points in east - em Montana, central Kansas, and north - central Texas in Fig. 6. Other studies support this finding. Stockton and Meko (1983) reconstructed annual precipitation for four regions flanking the Great Plains (centered in 2696 Vol. 79, No. 12, December 1998