Multiyear La Niña events and persistent drought in the contiguous United States



[1] La Niña events typically bring dry conditions to the southwestern United States. Recent La Niñas rarely exceed 2 years duration, but a new record of ENSO from a central Pacific coral reveals much longer La Niña anomalies in the 1800s. A La Niña event between 1855–63 coincides with prolonged drought across the western U.S. The spatial pattern of this drought correlates with that expected from La Niña during most of the La Niña event; land-surface feedbacks are implied by drought persistence and expansion. Earlier periods also show persistent La Niña-like drought patterns, further implicating Pacific anomalies and surface feedbacks in driving prolonged drought. An extended index of the Pacific Decadal Oscillation suggests that extratropical influences would have reinforced drought in the 1860s and 1890s but weakened it during the La Niña of the 1880s.

1. Introduction

[2] Many societal and natural systems are vulnerable to prolonged drought, including food production, water supplies, ecosystem health, and disease vectors. In the United States, even short-lived droughts have exacted large economic costs [Riebsame et al., 1991]. Multiyear droughts in the 1930s and 1950s caused tremendous disruption on social, agricultural, ecological, and economic fronts. Such droughts have recurred about twice per century in the past 400 years [Woodhouse and Overpeck, 1998]. The causes of drought and the feedbacks that enable its persistence are not understood sufficiently to explain multiyear droughts in the U.S. historical record. Here, we use a new record of El Niño/Southern Oscillation (ENSO) [Urban et al., 2000], a gridded reconstruction of Palmer Drought Severity Index (PDSI; [Cook et al., 1999]), and an extended index of the Pacific Decadal Oscillation (PDO) to explore links between multiyear U.S. droughts and Pacific ocean variability.

[3] Circulation features associated with U.S. drought include anomalous subsidence, reduced moisture transport, altered storm tracks, and an upper-level anticyclone near or upstream from the drought region [Namias, 1955; Dole, 2001]. Anomalous anticyclones can develop as wavetrains initiated by tropical or midlatitude Pacific SST anomalies [Trenberth et al., 1988; Ting and Wang, 1997]. Decreased evapotranspiration reduces local moisture availability and increases surface temperatures, prolonging drought [Namias, 1991; Lyon and Dole, 1995]. Over several dry years, vegetation death can allow soil mobilization that inhibits regrowth [Mangan et al., 2002].

[4] A typical La Niña produces unusually dry conditions in the southwest, extending into the Great Plains ( Drought in 1988 has been linked to coincident La Niña conditions [Trenberth et al., 1988]. Hydrologic impacts in the U.S. southwest are among the more reliable teleconnections of ENSO [Cole and Cook, 1998], and the impact is more consistent for La Niña than El Niño [Cayan et al., 1999]. Individual La Niña events vary, however, in the penetration of drought into adjacent regions. Extratropical Pacific SST anomalies may enhance or dampen ENSO's influence [Gershunov and Barnett, 1998].

[5] La Niña events in the 20th century are typically short-lived (≤2 years). A new record of ENSO from a central Pacific coral reveals prolonged (8–9 year) cool anomalies in the mid-late 19th century [Urban et al., 2000]. For this analysis, we removed multidecadal and longer variations from this record (Figure 1). La Niña conditions extend from 1855–1863 and from 1879–1888 (also in the original coral record). Such prolonged La Niñas do not occur in the 20th century. These decadal anomalies are almost certainly real features of ENSO. They occur in the Niño3.4 index and in other coral records from ENSO-sensitive regions (J. E. Cole, A widespread decadal oscillation in tropical climate during the 19th century, ms. in revision). The Maiana record correlates with the Multivariate ENSO Index [Wolter and Timlin, 1998] at r = −0.81.

Figure 1.

Record A shows a normalized and detrended history of ENSO variability from a central Pacific coral record [Urban et al., 2000]. Prolonged 19th-century La Niña events are shaded. Record B is the normalized index of drought area (DAI) in the coterminous U.S. [Cook et al., 1997, 1999]; horizontal line denotes the mean. The index is the number of 2° × 3° grid cells with PDSI ≤ −1, out of 154. Record C is the normalized PDSI in the 10 grid squares most strongly correlated to ENSO. Multiyear droughts (when 5-year mean values are above average for ≥6 consecutive years) are shaded. Record D shows correlation coefficients between the pattern of drought in a given year and the pattern of drought typical of 20th century La Niña events. Shading indicates coefficients between 0.25 and −0.25; values beyond these levels are significant at 95 to 99.9% depending on spatial autocorrelation. For all records, the heavy line is a 5-year running mean.

2. Results: Persistent Drought and La Niña

[6] We identify persistent drought using both a continental drought area index (DAI) [Cook et al., 1997] and a regional drought index covering the most ENSO-sensitive areas. In our comparison with the ENSO record, we focus on persistent droughts - intervals when the 5-year smoothed indices are anomalous for ≥6 consecutive years (Table 1). We use annual records to define the duration of these droughts.

Table 1. Multiyear Droughts (1700–1978) and Correlation with the La Niña Drought Pattern
DecadeDuration in DAIDuration in regional indexPattern correlationa
  • a

    Bold type indicates positive correlation significant at the 99% level after a reduction in the degrees of freedom based on Moran's I statistic of spatial autocorrelation [Sawada, 1999].


2.1. Diagnosis of the 1860s Example

[7] A persistent drought near 1860 coincides with a prolonged La Niña event in the coral record. This drought has been identified in several studies as particularly significant. In Texas tree-ring records, the decade centered at 1860 is the driest of the 1698–1980 record [Stahle and Cleaveland, 1988]. In tree-ring records bordering the Great Plains, the 1860s drought exceeds the 1930s Dust Bowl drought in multiyear intensity [Meko, 1992]. Historical accounts from the Great Plains and southwest note blowing sand and expanded dune fields at that time [Muhs and Holliday, 1995].

[8] The spatial pattern of drought varies from year to year during La Niña conditions. The demise of El Niño conditions in 1855 corresponds to the onset of widespread drought in the DAI but not the regional index. Yearly maps of reconstructed PDSI ( show that in 1856-7, drought is centered in southern California and the midwest, straddling the southwest. Although atypical of 20th-century La Niña events, this pattern occurs in the La Niña years of 1928–9 and 1971 and may reflect La Niña's influence in 1856–7. In 1858 the Pacific warms to normal, and the southwest is wet. From 1859–1862, La Niña conditions resume and drought is centered in Texas/New Mexico, extending into the midwest. Between 1863–5, drought expands across the western U.S., even as the Pacific returns to a normal and then warm state. The expansion and persistence of this drought suggest that initial dry years triggered land-surface feedbacks that may have reduced moisture availability and altered circulation.

2.2. Earlier Intervals

[9] Events in the DAI nearly always co-occur with events in the ENSO-sensitive region, implicating La Niña as a major cause of large-scale U.S. drought. Droughts that coincide exactly in these records occurred in 1703–09 and 1818–24. Those that partly coincide occur in the 1730s, 1750s, 1860s, 1890s, and 1950s; for these events, anomalies in the DAI often lag those in the regional index, suggesting that La Niña can trigger extensive dry periods and additional feedbacks can maintain them. In the 1780s, a drought in the ENSO-sensitive region was not seen in the DAI, and the ENSO-sensitive region did not experience prolonged drought during DAI events of the 1840s and the 1930s.

[10] The pattern of a drought may indicate its origin [Fye et al., 2002]. Composite maps of drought during recent La Niña years reveal a characteristic pattern (Figure 2) that may help identify past La Niña-related droughts. Using the 300-year PDSI reconstruction [Cook et al., 1999], we map patterns of past multiyear drought (Figure 3) and calculate their correlation to the 20th-century pattern (instrumental and reconstructed) of La Niña drought (Table 1). Yearly correlation values (Figure 1) indicate whether significant multiyear correlations reflect a persistent relationship. Strongest multiyear correlations occur during the 1820s, 1860s, and 1950s. Although correlations are significant, the modern La Niña pattern explains only <30–50% of variance in the spatial pattern of past drought, probably due to the inherent variability of extratropical ENSO impacts [Hoerling and Kumar, 1997]. High correlation during the 1950s drought is partly because the years that define the La Niña pattern include two between 1949–56; correlations remain significant if those years are removed from the composite.

Figure 2.

Average drought pattern during 20th century La Niña years, using (a) instrumental and (b) reconstructed PDSI data. We selected La Niña years based on anomalies in both Niño 3.4 SST and Darwin sea level pressure (−0.8°C and −1 mbar, respectively, in 3-month smoothed records). Years thus identified are 1910, 1917, 1925, 1950, 1956, 1971, 1974, and 1985. Other identification methods [Gershunov and Barnett, 1998] yield similar patterns. The reconstructed drought series ends in 1978; composite maps exclude 1985.

Figure 3.

Mapped patterns of reconstructed PDSI for the intervals identified in Table 1.

3. Discussion

3.1. Implications for Earlier Anomalies

[11] Spatial patterns of persistent droughts over the past 300 years implicate Pacific variability in many cases. Droughts most similar to the La Niña pattern occurred in the 1700s, 1820s, 1860s, 1890s, and 1950s, although in the 1700s and 1890s the correlation with La Niña is driven by a few individual years. The pattern of the 1930s Dust Bowl contrasts strongly with these southwest-focused droughts, covering mainly the northern half of the U.S. The 1730s drought pattern appears similar to a La Niña pattern in orientation, but displaced westward (as are droughts in the 1700s and 1890s); the 1750s drought impacts regions common to both the Dust Bowl and La Niña patterns. Independent records of ENSO cannot yet confirm whether La Niña events occurred during prolonged droughts between 1700–1840.

[12] A subtle difference between the 20th century La Niña pattern and earlier persistent droughts is the westward displacement of the anomalies in earlier periods; earlier droughts are typically centered in Arizona/California, whereas the 20th-century La Niña pattern centers on Texas/New Mexico. Confirming that this change reflects a systematic shift in the ENSO teleconnection pattern requires longer ENSO records to identify La Niña conditions unambiguously. This shift may have reversed in recent decades [Cole and Cook, 1998].

3.2. The Role of the North Pacific

[13] The La Niña-SW drought connection is not infallible. A persistent La Niña in 1879–1885 does not correspond to prolonged drought, nor does 1890s drought co-occur with La Niña. ENSO impacts are modulated by North Pacific variability, often described in terms of SST patterns (the PDO) [Mantua et al., 1997]. The original PDO index ends in 1900. To examine the PDO's influence on 19th century drought-La Niña connections, we extend the PDO index to 1856 using an interpolated SST product [Kaplan et al., 1998]. The extended PDO agrees well (r = 0.81) with the original (Figure 4).

Figure 4.

History of the Pacific Decadal Oscillation (PDO), with original values [Mantua et al., 1997] in the light line and the new extended PDO in black.

[14] When the PDO is negative, La Niña teleconnections tend to be strong and stable [Gershunov and Barnett, 1998]. Persistent negative PDO values from 1858–1864 would enhance the impact of the concurrent La Niña. However, during the La Niña of 1879–85 the PDO oscillated between positive and negative values. From 1879–1883, the PDO was slightly negative and the southwest U.S. experienced moderate and/or localized drought ( In 1884, the PDO switched to positive values; despite the persistence of La Niña conditions until 1888, the southwest drought ended. A return to negative PDO values in the early 1890s coincides with the worst drought years of that interval, although La Niña conditions were weak or absent.

4. Conclusions

[15] Decadal variability in the tropical Pacific, enhanced by extratropical conditions, likely contributed to unusually strong drought during the 1860s and perhaps other intervals. Between 1818–1824, a distinctive multiyear drought pattern implies concurrent prolonged La Niña conditions. La Niña appears to trigger drought which can then outlive the La Niña due to additional feedbacks. The growing network of proxy records, in combination with model experiments and process studies of surface feedbacks, can help us understand the causes of multiyear droughts in the past few centuries; a greater challenge is to understand the multidecadal megadroughts that appear unambiguously in paleoclimatic records [Woodhouse and Overpeck, 1998; Stahle et al., 2000]. The rapid growth of populations and economies in semi-arid regions worldwide highlights the need for a predictive understanding of persistent drought.


[16] This work was supported by NSF's Earth System History and Climate Dynamics programs (JEC and JTO) and NOAA's Office of Global Programs (ERC). We are grateful to Connie Woodhouse and Dave Stahle for discussion and comments on the paper, to Frank Urban for his work on the Maiana coral dataset, and to Mike Evans for additional input. Comments by anonymous reviewers greatly improved this manuscript. Thanks to Michael Sawada for the Rook's Case software to calculate spatial autocorrelation statistics. This is LDEO contribution 6262.