1 AUGUST 2005
WATSON AND LUCKMAN
2847
Spatial Patterns of Preinstrumental Moisture Variability in the Southern Canadian Cordillera EMMA WATSON Climate Research Branch, Meteorological Service of Canada, Environment Canada, Downsview, Ontario, Canada
BRIAN H. LUCKMAN Department of Geography, University of Western Ontario, London, Ontario, Canada (Manuscript received 7 May 2004, in final form 23 October 2004) ABSTRACT Extreme wet and dry intervals of the last 350 yr in the Canadian Cordillera and adjacent United States are examined using a network of 25 tree-ring-based precipitation and Palmer Drought Severity Index (PDSI) reconstructions. Reconstructed twentieth-century-mapped patterns compare well with patterns based on the instrumental records at both annual and decadal scales. During the most extreme events, dry conditions occurred over the entire area. The longest widespread drought in the last 350 yr occurred from 1917 to 1941. Shorter intervals of more severely dry conditions occurred in the early 1720s, 1750s, 1790s, 1860s–70s, and the 1890s. Many of the driest individual years and most extreme dry periods of ⬍7 yr are reconstructed for the eighteenth century. The longest, wettest periods identified by these reconstructions occurred in the early twentieth century. In agreement with published studies that explore links between instrumental precipitation records from the region and conditions in the Pacific Ocean, the reconstructed records show that drier (wetter)-than-normal conditions are associated with El Niño (La Niña) events and the positive (negative) phase of the Pacific decadal oscillation (PDO).
1. Introduction The majority of dendroclimatic work conducted in the southern Canadian Cordillera has focused on generating estimates for individual climatic time series: either for a single meteorological station (e.g., Watson and Luckman 2001a) or a regionally representative time series (e.g., Luckman et al. 1997; Wilson and Luckman 2003). Although these studies have greatly expanded our knowledge of climatic variability at sites within the region, they have not systematically explored variability across the region. Climate reconstructions for individual records contain local-, regional- and large-scale signals that cannot be differentiated, or verified, without the broader view and replication that a network of sites affords. The recent development of a gridded network of Palmer Drought Severity Index (PDSI) reconstructions for the contiguous United
Corresponding author address: Dr. Emma Watson, Climate Research Branch, Meteorological Service of Canada, 4905 Dufferin Street, Downsview, ON, M3H 5T4, Canada. E-mail:
[email protected]
© 2005 American Meteorological Society
States (Cook et al. 1996, 1999) has demonstrated the tremendous potential for tree-ring-based reconstructions to document spatial as well as temporal patterns of preinstrumental climate variability. Reconstruction of such patterns allows us to compare the magnitude and spatial extent of documented extreme events (e.g., the 1920s–1930s drought) with those in the preinstrumental period. This type of information cannot be elicited from individual station reconstructions and is highly significant because of the relatively greater economic and social costs of widespread prolonged drought vis-à-vis shorter, more localized severe events. Much of the southern Canadian Cordillera and the neighboring prairies lie in the rainshadow of adjacent mountains and consequently have low mean annual precipitation totals (often less than 400 mm). The spatial and temporal variability of precipitation is therefore a significant control of many economic activities and an important conditioner for forest fire activity within the region. The extreme forest fire season of 2003 in the British Columbia interior was preceded by three years of drought and cost the province an estimated 700 million dollars, destroyed over 300 homes,
2848
JOURNAL OF CLIMATE
and led to the evacuation of over 45 000 people (Filmon 2004). Watson and Luckman (2001a, 2004) have developed tree-ring-based reconstructions of annual precipitation for 13 meteorological stations across this region. These data, combined with precipitation/PDSI reconstructions for the Rocky Mountain foothills, the Canadian Prairies, and adjacent United States provide the first dataset that can be used to explore spatial aspects of preinstrumental precipitation in western Canada. This study complements earlier research into temporal characteristics of dry and wet events in the region using these data (Watson and Luckman 2004). In this paper, we present maps of the spatial extent of dry and wet periods in the southern Canadian Cordillera over the past 350 yr. The periods identified during the twentieth century can be compared with those in the preinstrumental period. This allows us to determine whether these earlier events were more extensive and/ or severe than those identified in the instrumental record (e.g., the infamous drought of the 1920s–30s). It also provides data about the past frequency of widespread dry and wet events that persist over different time scales across the region. To improve our understanding of the causes of widespread events, we provide a preliminary discussion of their possible association with conditions in the Pacific Ocean.
2. Datasets and reconstruction quality a. Datasets used The primary dataset used in this analysis consists of 13 annual (prior July to current June) precipitation reconstructions developed from a network of Pseudotsuga menziesii (Douglas fir) and Pinus ponderosa (ponderosa pine) tree-ring chronologies sampled at low elevation sites in the southern Canadian Cordillera (Watson and Luckman 2001b, 2004). In addition, to expand spatial coverage, we utilize data from 10 adjacent grid points from the June–August PDSI estimates in the northwestern United States (Cook et al. 1999) plus annual (prior August to current June) precipitation reconstructions for the Rocky Mountain foothills (Case and MacDonald 1995) and Canadian Prairies (Sauchyn and Beaudoin 1998; Fig. 1). Although these sites are unevenly distributed across the region (Fig. 1), Watson and Luckman (2004) demonstrate that the reconstructions from the southern Canadian Cordillera correlate significantly with gridded instrumental precipitation data over the region and define coherent spatial patterns. This indicates that these reconstructions capture regional precipitation variability and are therefore appropriate for use in this study.
VOLUME 18
FIG. 1. Location of the tree-ring reconstructions used in this study. Annual (Jul–Jun) precipitation reconstructions developed by Watson and Luckman (2004) are PG ⫽ Prince George, BC ⫽ Big Creek, WL ⫽ Williams Lake, JA ⫽ Jasper, NT ⫽ North Thompson, LI ⫽ Lillooet, LY ⫽ Lytton, WE ⫽ Westwold, BA ⫽ Banff, SU ⫽ Summerland, OL ⫽ Oliver, CR ⫽ Cranbrook, and WA ⫽ Waterton. Here, CM is an annual precipitation reconstruction (Aug–Jul for a combined Fort Macleod and Calgary record) using limber pine (Case and MacDonald 1995). SAU is an annual precipitation reconstruction (Aug–Jul) for Maple Creek, Saskatchewan, using white spruce from the Cypress Hills (Sauchyn and Beaudoin 1998). The grid points (GPs) are gridded Jun–Aug PDSI reconstructions from Cook et al. (1999). The shaded area in Canada represents the approximate location of the Palliser Triangle, which is an extension of the U.S. northern Great Plains (also shaded).
Although proxy records of two different variables (precipitation and PDSI) are used in this study, the two moisture measurements are comparable. The PDSI (Palmer 1965) is a measure of meteorological drought that incorporates precipitation, potential evapotranspiration, and antecedent soil moisture and runoff (Oladipo 1985). Monthly PDSI values have considerable month-to-month persistence built into them (Hu and Wilson 2000) and therefore represent conditions over a much longer time period than monthly precipitation. Annual instrumental precipitation totals (summer– summer) correlate highly with summer PDSI values in the southern Canadian Cordillera (Watson and Luckman 2002). For example, the mean correlation between August and July precipitation at 17 stations in the area and the June–August (JJA) PDSI (the period used by Cook et al. 1999) is 0.79. As others report similar results (e.g., Stahle et al. 2000) we consider these two parameters to be comparable.1 1
Few instrumental PDSI records are available for the southern Canadian Cordillera and therefore preference was given to reconstructing precipitation variables (Watson and Luckman 2002).
1 AUGUST 2005
WATSON AND LUCKMAN
b. Comparing instrumental and reconstructed precipitation patterns The individual reconstruction models pass conventional verification tests that evaluate the year-to-year performance and levels of explained variance in the time series over several intervals (Watson and Luckman 2004; Cook et al. 1999; Sauchyn and Beaudoin 1998; Case and MacDonald 1995). However, these tests do not evaluate the collective ability of the reconstructions to reconstruct spatial patterns. The degree of similarity between spatial patterns may be assessed visually and by calculating correlation and congruence coefficients (Richman 1986; Cook et al. 1999). Congruence coefficients provide more information on the similarity of the magnitude of the anomalies in each annual map pair than correlation coefficients that provide a metric of how well the annual map patterns covary (Richman 1986; Cook et al. 1999). The mean correlation and congruence coefficients over the interval 1920–94 are 0.60 and 0.67, respectively (Fig. 2). This difference is not unexpected, as congruence coefficients are biased to higher values than the correlation coefficient (Richman 1986; Cook et al. 1999). These values are similar to those identified by Cook et al. (1999) between instrumental and reconstructed patterns of summer PDSI over a much larger region. These results indicate that the reconstructions consistently capture large-scale annual spatial patterns providing a convincing verification of both the signal at individual sites and its spatial representativeness. The ability to replicate the annual patterns is, in and of itself, a verification of the skill of the individual reconstructions. Comparison of the reconstructed and instrumental annual patterns (Fig. 2) indicate that only about 10% (around 8 of 75, depending on the criterion used) of the years are poorly modeled. An examination of the instrumental and reconstructed patterns for these years (Fig. 3) is instructive. The annual contour maps in Fig. 3 show reconstructed and instrumental annual precipitation/PDSI values expressed in standard deviation units based on the period 1920–94. In all cases, the reconstructed patterns include some points that are the incorrect sign (i.e., a reflection of local climatic or nonclimatic anomalies) and others that are the correct sign but underestimate (or overestimate) the magnitude of deviation. However, if one were to coarsely summarize the entire region as being either wet or dry (or half wet and half dry), the reconstructions perform well even for these poor years. When the annual maps are composited over longer intervals (Fig. 4), local anomalies are reduced by the averaging process making longer-term patterns more comparable than those for individual
2849
years. Mean correlation and congruence coefficients calculated between the pairs of decadal maps (similar to the maps of widespread anomalies presented later in this paper) are higher than for individual years (0.78 and 0.79; Fig. 4).
3. Results a. Evaluation of the relative severity of prolonged dry and wet periods The magnitude of prolonged dry and wet intervals can be defined by ranking the mean standardized anomalies for all 25 reconstructions (Tables 1a and 2a) or the 13 records from the southern Canadian Cordillera (Tables 1b and 2b) for equivalent intervals of time. Only 1 (entire area, Table 1a) or 3 (13 site network, Table 1b) of the 10 most extreme dry individual years occur in the twentieth century compared with 5 (Table 2a) or 3 (Table 2b) of the 10 wettest years. These results suggest that the twentieth-century instrumental record does not contain many examples of the most extreme dry years experienced in this region in the last 350 yr. At the decadal time scale, a dry period originating in the first quarter of the eighteenth century (⬃1717) is identified as well as a period in the 1790s (Table 1a only, it is identified in shorter intervals in Table 1b) and the early twentieth-century drought. Shorter dry intervals (i.e., ⬍7 yr) are identified for the late 1750s to early 1760s, and the late 1860s through early 1870s (Tables 1a and 1b). The 3-yr mean for the period 1661–63 is identified as dry for the southern cordillera alone (Table 1b). Several of these shorter intervals (those in the early and mid-1700s) are also considerably more severe than equivalent periods during the early twentieth century. At longer time scales (⬎10 yr means), the documented drought that occurred in the early twentiethcentury dominates (Table 1). The reconstructions indicate that it is clearly the longest, most severe dry spell to have affected the region in the past 350 yr. Within this drought, the 3-yr interval 1929–31 was particularly dry (Tables 1a and 1b). The next most severe 20-yr periods in the entire region in the past 350 yr are 1717– 36 (mean ⫺0. 25, ranked 14th), 1781–1800 (mean ⫺0.18, ranked 21st), and 1840–59 (mean ⫺0.15, ranked 29th; not shown in Table 1a). The strongest, most prolonged wet interval appears to have occurred in the early part of the twentieth century (⬃1897–1916: Tables 2a and 2b). There are some differences in the order of severity of other prolonged wet intervals between the entire region (Table 2a) and the southern Canadian Cordillera (Table 2b). In the
2850
JOURNAL OF CLIMATE
FIG. 2. Measures of agreement between the reconstructed and instrumental annual maps. (a) The number of stations available in each pair of annual maps. (b) Correlations between the instrumental and reconstructed maps in each year. Years with nonsignificant correlation coefficients are labeled prior to 1980, after which correlations become a less reliable measure because of the decreased number of observations. Note that the significance levels displayed do not account for any spatial autocorrelation that may exist in the dataset. They are only a guide to identify poorly modeled years. (c) Congruence coefficients for the same data. There is no theoretical sampling distribution for testing the significance of congruence coefficients (Richman 1986).
VOLUME 18
1 AUGUST 2005
WATSON AND LUCKMAN
2851
FIG. 3. A comparison of the instrumental and reconstructed precipitation/PDSI anomalies in the poorly modeled years identified in Fig. 2b. Data are expressed as standard deviations from the 1920–94 mean (or, for shorter records, the longest possible portion of this interval).
series for the entire region (Table 2a), anomalous wet periods occurred in the first quarter of the twentieth century, 1900–19 and in the late seventeenth century (⬃1666–85). In the series for the southern Canadian Cordillera, the second, third, and fourth wettest 20-yr periods include the mid-twentieth century (⬃1947–66), the period 1801–20, and the late seventeenth century (as seen in the regional series). At shorter intervals, the latter part of the 1870s to the mid-1880s is also reconstructed as being particularly wet.
b. Evaluation of mapped patterns In the previous section, reconstructed dry and wet events were ranked according to their severity across the region for a number of time scales. This approach excludes information on their spatial extent and less severe but more extensive droughts may not be identified. It is also difficult to define the most appropriate length of period to identify dry and wet events as they overlap at a number of time scales (Tables 1 and 2). In
2852
JOURNAL OF CLIMATE
VOLUME 18
FIG. 4. A comparison of patterns of instrumental and reconstructed precipitation/PDSI anomalies by decade from 1920 to 1989. Here, r and cc refer to the correlation and congruence coefficients between each pair of decadal maps. Data are expressed as standard deviations from the 1920–94 mean.
the following sections, we present maps of dry and wet events (defined according to their spatial extent) at three time scales: annual events, 5–10-yr periods and intervals ⬎10 yr in length.
1) ANNUAL
MAPS
Patterns of precipitation anomalies may be mapped over the network for each year from about 1640 to 1996. Prior to 1640, there are ⬍10 records available and the contoured patterns are unduly influenced by the distribution of the records. Between 1640 and 1996, the number of records varies between 9 and 25 because of the different lengths of the individual reconstructions.
Conditions during two of the most extreme dry and wet years reconstructed (Fig. 5a; Tables 1a and 2a) display large areas where precipitation/PDSI is more than one standard deviation from the mean and roughly one quarter of the values are more than two standard deviations from the mean. These diagrams indicate that for individual years strong anomalies of the same sign can occur across the entire region. Figures 5b and 5c show less extreme sequences of consecutive years that are highly significant in the history of settlement of the Canadian west. The Palliser Triangle is a semiarid, drought-prone region of the Canadian Prairies located in southeast Alberta and south-
1 AUGUST 2005
2853
WATSON AND LUCKMAN
TABLE 1. The 10 lowest mean standardized anomalies ranked over different intervals for (a) the entire mapped area and (b) the southern Canadian Cordillera from 1650 to 1996. The number of reconstructions (N) used for area (a) ranges from 25 (1978) to 10 (1650). For the southern Cordillera (b) N decreases from 13 in 1994 to 6 in 1650. All reconstructions were converted to standardized anomalies relative to the 1831–1978 common period. The bold font identifies the first occurrence of a period within each interval. Italicized periods overlap with a period already identified in that interval. Gray shading identifies periods that overlap with the persistent (20-yr mean) drought identified for the years 1917–36. (a)
1 Year 1 2 3 4 5 6 7 8 9 10 (14)* (21)*
(b)
1869 1721 1756 1657 1741 1931 1800 1677 1717 1794
⫺1.46 ⫺1.43 ⫺1.41 ⫺1.37 ⫺1.30 ⫺1.24 ⫺1.24 ⫺1.24 ⫺1.22 ⫺1.10
1 Year 1 2 3 4 5 6 7 8 9 10 (13)* (31)*
1869 1759 1929 1931 1720 1760 1741 1757 1922 1663
⫺1.78 ⫺1.34 ⫺1.33 ⫺1.25 ⫺1.18 ⫺1.18 ⫺1.15 ⫺1.06 ⫺1.04 ⫺1.04
3 Year 1719–1721 1929–1931 1717–1719 1756–1758 1739–1741 1718–1720 1869–1871 1918–1920 1924–1926 1720–1722
5 Year ⫺0.97 ⫺0.93 ⫺0.89 ⫺0.84 ⫺0.83 ⫺0.80 ⴑ0.79 ⫺0.74 ⫺0.72 ⫺0.71
1717–1721 1756–1760 1718–1722 1922–1926 1716–1720 1794–1798 1796–1800 1918–1922 1793–1797 1755–1759
3 Year 1929–1931 1758–1760 1757–1759 1869–1871 1719–1721 1759–1761 1924–1926 1661–1663 1922–1924 1930–1932
10 Year ⫺1.01 ⴑ0.83 ⫺0.71 ⫺0.68 ⫺0.65 ⫺0.65 ⫺0.65 ⫺0.64 ⫺0.62 ⫺0.61
1917–1926 1922–1931 1717–1726 1791–1800 1918–1927 1929–1938 1928–1937 1716–1725 1930–1939 1923–1932
5 Year ⫺1.06 ⫺1.03 ⫺0.99 ⫺0.98 ⫺0.81 ⫺0.79 ⫺0.79 ⴑ0.77 ⫺0.74 ⫺0.74
1756–1760 1757–1761 1922–1926 1929–1933 1918–1922 1869–1873 1868–1872 1928–1932 1921–1925 1717–1721
20 Year ⫺0.58 ⫺0.56 ⴑ0.55 ⴑ0.53 ⫺0.52 ⫺0.52 ⫺0.50 ⫺0.49 ⫺0.48 ⫺0.47
10 Year ⴑ0.85 ⫺0.80 ⫺0.78 ⫺0.75 ⫺0.71 ⴑ0.70 ⫺0.69 ⫺0.68 ⫺0.67 ⫺0.66
1922–1931 1917–1926 1918–1927 1923–1932 1921–1930 1920–1929 1924–1933 1717–1726 1919–1928 1925–1934
1918–1937 1917–1936 1919–1938 1922–1941 1920–1939 1921–1940 1916–1935 1915–1934 1923–1942 1924–1943 1717–1736 1781–1800
ⴑ0.51 ⫺0.50 ⫺0.46 ⫺0.46 ⫺0.45 ⫺0.45 ⫺0.39 ⫺0.35 ⫺0.35 ⫺0.32 ⴑ0.25 ⴑ0.19
20 Year ⫺0.70 ⫺0.67 ⫺0.64 ⫺0.63 ⫺0.63 ⫺0.62 ⫺0.60 ⴑ0.59 ⫺0.55 ⫺0.55
1917–1936 1918–1937 1919–1938 1916–1935 1920–1939 1921–1940 1915–1934 1922–1941 1914–1943 1913–1942 1712–1731 1852–1871
ⴑ0.56 ⫺0.53 ⫺0.50 ⫺0.48 ⫺0.47 ⫺0.47 ⫺0.47 ⫺0.46 ⫺0.42 ⫺0.38 ⴑ0.34 ⴑ0.23
* Identifies the two most highly ranked nonoverlapping 20-yr periods.
west Saskatchewan. It is named after the British explorer Captain John Palliser who visited the region in the late 1850s and said that it was too dry to be suitable for agricultural settlement (Fig. 5b; Koshida 1992; Palliser 1859). In a later expedition during the 1870s, the botanist John Macoun found the region to be suitable for agriculture based on the comparatively wet conditions that prevailed during that interval (Fig. 5c; Nemanishen 1998; Koshida 1992). His findings encouraged settlement of the area that we now know receives an average of ⬍350 mm of precipitation per year but produces 75% of Canadian wheat (Koshida 1992). Agreement between these documentary sources and the treering-based reconstructions provides mutual verification while demonstrating the important context that these longer records provide for evaluating more recent instrumental climate records. Although it is important to examine the magnitude of departures in widespread patterns that occur in particular years (e.g., for water resource planning), it is perhaps more important to study widespread patterns that
persisted for several years. The examples presented in Fig. 5 are not isolated extreme years but occur within sequences where the pattern of departures may persist for several years and thereby have a more pronounced impact than single year extreme events. In the rest of this paper we present a series of maps that explore longer-term, widespread dry and wet anomalies across the region.
2) WIDESPREAD
DRY AND WET INTERVALS
Major widespread dry and wet periods reconstructed for the southern Canadian Cordillera over the past ⬃350 yr are presented in Figs. 6 and 7. These periods are defined as intervals longer than 4 yr when ⬎50% (⬎60% before 1700 when n ⬍ 8) of the 13 annual precipitation reconstructions developed for British Columbia and Alberta are below (dry) or above (wet) their 1831–1978 average (the period common to all 25 reconstructions). In some cases, single years within the interval fall below this 50% criterion. These intervals
2854
JOURNAL OF CLIMATE
VOLUME 18
TABLE 2. The 10 highest mean standardized anomalies ranked over different intervals. For explanation see Table 1. (a)
1 Year 1 2 3 4 5 6 7 8 9 10
1715 1672 1697 1750 1996 1775 1913 1981 1916 1942
1 2 3 4 5 6 7 8 9 10
1697 1672 1752 1996 1900 1789 1819 1670 1879 1981
(b)
1.38 1.36 1.32 1.28 1.22 1.17 1.16 1.16 1.16 1.15
1 Year 2.04 1.67 1.46 1.42 1.39 1.34 1.24 1.23 1.21 1.20
3 Year 1879–1881 1877–1879 1750–1752 1994–1996 1752–1754 1666–1668 1992–1994 1878–1880 1899–1902 1911–1913
5 Year 0.86 0.81 0.81 0.78 0.75 0.73 0.73 0.69 0.68 0.68
1990–1994 1877–1881 1750–1754 1912–1916 1993–1997 1876–1880 1992–1996 1666–1670 1749–1753 1875–1879
1.33 1.22 1.07 1.00 0.96 0.93 0.91 0.88 0.85 0.84
1694–1698 1693–1697 1696–1701 1695–1699 1900–1904 1990–1994 1877–1881 1750–1754 1899–1903 1672–1676
3 Year 1696–1698 1695–1697 1697–1699 1900–1902 1879–1981 1670–1672 1750–1752 1805–1807 1804–1806 1752–1754
10 Year 0.78 0.77 0.73 0.72 0.72 0.70 0.69 0.67 0.62 0.62
1989–1988 1666–1675 1907–1916 1667–1676 1873–1882 1664–1673 1988–1997 1876–1885 1875–1884 1665–1674
1.10 1.01 1.00 0.97 0.89 0.80 0.77 0.76 0.75 0.72
1691–1700 1693–1702 1690–1699 1666–1675 1692–1701 1689–1698 1665–1674 1664–1673 1667–1676 1694–1703
5 Year
are defined using the method outlined in Watson and Luckman (2004). The aggregate maps produced reduce the localized anomalies created by poor estimates for individual years in specific reconstructions. The standardized anomalies on these maps are defined by the standard deviations of the annual series over the 1831–1978 common period. However, averaging reduces the absolute magnitude of these anomalies that display a reduced range of departures from those seen in the annual maps (e.g., Figure 5; Table 1). For this reason, the absolute magnitudes of the anomalies in the annual (Figs. 3 and 5) and composite maps (Figs. 4 and 6–8) are not directly comparable as they express anomalies averaged over different time intervals. The absolute values are compared for specific intervals in Tables 1 and 2. The maps of the extended dry and wet intervals include data from precipitation and PDSI reconstructions in adjacent areas that were not used in defining these events. The coincident occurrence of pronounced moisture anomalies in reconstructions from adjacent areas verifies the significance of the periods identified using the smaller 13-reconstruction dataset. If the spatial scale was broadened and all 25 reconstructions were used to define extreme periods, fewer intervals would be identified because it is less common to find droughts that affect the entire region. Smaller-scale dry/wet events unique to the southern Canadian Cordillera would not be as readily identified. Therefore these ad-
20 Year 0.54 0.52 0.49 0.47 0.46 0.45 0.45 0.45 0.44 0.43
1897–1916 1898–1917 1899–1918 1978–1997 1896–1915 1979–1998 1900–1919 1977–1996 1666–1685 1894–1913
0.70 0.66 0.65 0.65 0.64 0.64 0.62 0.60 0.59 0.57
1897–1916 1947–1966 1898–1917 1801–1820 1948–1967 1664–1683 1663–1682 1662–1681 1802–1821 1665–1684
10 Year
0.40 0.40 0.35 0.34 0.32 0.30 0.28 0.28 0.27 0.27
20 Year 0.41 0.39 0.39 0.39 0.38 0.37 0.37 0.36 0.36 0.35
ditional data are included to provide more information about the spatial scale of the reconstructed anomalies. Maps of the dry and wet periods are arranged according to their length to enable direct comparison of periods of similar length. The patterns or the spatial scale of dry/ wet events of different lengths should be more directly comparable than the actual magnitude of departures.
(i) Dry periods ⬍10 yr in length The majority of the spatially extensive dry periods identified are less than 10 yr in length (Fig. 6). The most severe are in the mid-1750s, which coincides with dry conditions over much of the western United States, and the 1790s. On average ⬃60% of the 24 reconstructions that extend back to the 1790s drought have values more than half a standard deviation below their mean with particularly dry conditions reconstructed in the Rockies and the foothills. A streamflow reconstruction for the North Saskatchewan River, which flows eastward from this region, reconstructs very low flows during the 1790s (Case and MacDonald 2003). Wolfe et al. (2001) document a widespread period of dune activity in the Great Sand Hills region of the Palliser Triangle in Saskatchewan during the early 1800s. They attribute this activity to the dry conditions that prevailed throughout much of the 1700s and the extremely dry conditions of the 1790s [as based on precipitation reconstructions in Case and MacDonald (1995) and Sauchyn and Beaudoin (1998), also used in
1 AUGUST 2005
WATSON AND LUCKMAN
FIG. 5. Annual maps showing reconstructed moisture conditions over a number of intervals. (a) Two of the driest and wettest years reconstructed (see Tables 1 and 2); n is the number of reconstructions used to produce the map. The percentage of records that are more than 0.5 standard deviations (sd) from their 1831–1978 mean are listed. (b), (c) Selected years during the nineteenth century (refer to text for discussion). The maps show reconstructed conditions for each record expressed in standard deviation units from the mean of the period common to all 25 reconstructions (1831–1978).
2855
2856
JOURNAL OF CLIMATE
FIG. 6. Widespread dry periods in the southern Canadian Cordillera after 1650. The dry intervals are defined as periods when ⬎50% (⬎60% before 1700) of the 13 reconstructions developed by Watson and Luckman (2004; Fig. 1) show below average precipitation for ⬎4 consecutive years. Single years within some intervals fall below the 50% criterion. The z scores are derived using the mean and standard deviation calculated over the common period of all reconstructions (1831–1978). (a), (b) Dry intervals ⬍ and ⬎10 yr are plotted sequentially, and (c) subdivisions of the 1917–41 period are shown. The number of years in each period is given in brackets.
VOLUME 18
1 AUGUST 2005
WATSON AND LUCKMAN
FIG. 7. Widespread wet periods in the southern Canadian Cordillera after ⬃1650. For explanation see Fig. 6.
2857
2858
JOURNAL OF CLIMATE
this study]. The results presented here support this interpretation as the greatest concentration of dry periods identified in the past 350 yr occurs during the 1700s. In addition to the 1750s and 1790s, dry intervals are identified from 1701–08 and 1768–72. During both of these periods, most of the PDSI records from the United States show wetter than normal conditions. The 1868–75 drought is not particularly striking in terms of the magnitude of the standardized anomalies or its duration, but it appears to have had a strong impact on the landscape. In particular, the year 1869 was very dry (80% of the 25 reconstructions have values more than one standard deviation below the mean).2 The widespread dry conditions estimated for 1869 coincide with fire-scar evidence of an extreme fire year in the Cariboo Forest region of southern British Columbia (Daniels and Watson 2003; Gray et al. 2002) and with documented crop failure and grasshopper plagues on the Canadian Prairies in 1868 (Phillips 1990). The period 1889–97 was also very dry across the region with the driest conditions estimated for the eastern reconstructions. Major forest fires burned approximately 9% and 20% of Banff and Jasper National Parks in the Canadian Rockies in 1889 (Luckman 1998). Phillips (1990) reports that 9 yr of drought during the 1890s led to farm abandonment on the Canadian Prairies and severe drought also occurred in the U.S. Great Plains during the 1890s (Woodhouse and Overpeck 1998). The 1750s, 1790s, 1860s, and 1890s droughts are recorded in a tree-ring-based reconstruction of July PDSI developed for the Canadian Prairies by Sauchyn and Skinner (2001). The most recent short (i.e., ⬍10 yr), dry interval identified for the region occurred in the 1980s (1985– 89).3 This drought had severe impact on agriculture in western Canada and an estimated 10% of farmers and farm workers left the agricultural sector in 1988 (Phillips 1990). The year 1985 was one of the worst fire years on record in British Columbia and cost roughly 300 million dollars in timber losses and fire-fighting expenses (Phillips 1990).
(ii) Dry periods ⬎10 yr in length The mapped patterns in Fig. 6 show five discrete dry periods greater than 10 yr in length during the past 350 yr. The earliest reported here is 1641–53 during which time 8 of the 10 available records show below-normal
2
The reconstructed 1869 year is July 1868–June 1869. The PDSI reconstructions are not included in this map because they terminate in 1978. 3
VOLUME 18
conditions with the driest records located in British Columbia. Mean moisture conditions are reconstructed as below normal for all records during the 15-yr interval 1717–31. During the driest portion of this interval, 1717–21, reconstructed values are on average more than one standard deviation below the common period mean. Evidence for this drought is also indicated by below-normal streamflow estimates for the North Saskatchewan, South Saskatchewan, and Saskatchewan Rivers (Case and MacDonald 2003), and negative summer PDSI estimates on the Canadian Prairies for this interval (Sauchyn and Skinner 2001). Prolonged dry conditions are estimated for much of the study area over the years 1839–59. This drought corresponds with low streamflow estimates for the South Saskatchewan and Saskatchewan Rivers (Case and MacDonald 2003) and a drought in the western Great Plains of the United States from 1845 to 1856 that may be linked with the decline in the bison population of the area (Woodhouse et al. 2002). Fye et al. (2003) document drought over large areas of the western United States for much of this interval. Dry conditions are also reconstructed for the Canadian Prairies during the 1850s and 1860s (Sauchyn and Skinner 2001). The longest severe drought identified in the reconstructions occurred during the period 1917–41. All reconstructions show drier-than-normal conditions during this well-known dry interval. Analysis of the annual maps (not shown) reveals that the region of greatest dryness differs between the first and second half of the drought. During the 1920s, the most extreme dry conditions are seen for southern interior British Columbia whereas, during the 1930s, the driest conditions are seen in the southern part of the study region (Fig. 6c). However, the comparison of the instrumental and reconstructed decadal maps for the 1920s (Fig. 4) indicates that the severely dry area in the 1920s extended further south. Over the twentieth century, 7 of the 10 driest years4 in the instrumental records and 8 of those in the reconstructed records occur during this drought.
(iii) Wet periods Eleven regionally extensive wet periods dating from the late seventeenth century onward are displayed in Fig. 7. This diagram demonstrates that it is not uncommon to have wet periods that span the entire study
4
Calculated as the mean standard deviation from the 1920–94 mean for the set of 25 instrumental records and the set of 25 reconstructed records.
1 AUGUST 2005
WATSON AND LUCKMAN
region. Sample depth is limited (n ⫽ 10) during the earliest of these intervals (1664–68). All 23 reconstructions covering the 1749–55 interval show above-normal values averaging 0.58 standard deviations above the mean, with the greatest anomalies in the western part of the study region. Wetter-than-normal conditions are reconstructed for almost all of the records during the interval 1778–82. The final two short wet periods occur in the late twentieth century: 1980–84 and the more widespread wet period in 1990–94. Both of these periods are also seen in the instrumental records from the region. Six wet periods greater than 10 yr in length are identified in the reconstructions. The 12-yr period 1689– 1700 is wet across much of the region with the exception of a few records in the southern interior of British Columbia and the most easterly sites. The next four wet periods span parts of the nineteenth century. Notable among these is the period 1819–30, which coincides with the largest flood reconstructed for the Red River in Manitoba in 1826 (St. George and Nielsen 2003). Although the Red River drainage basin is east of the study region, these wet conditions could have extended across the prairies as wet conditions are reconstructed across much of the western United States and the U.S. Great Plains for part of this interval (Fye et al. 2003). The final two wet periods (1898–1916 and 1942–60) bracket the 1920s–1930s drought. The wet period in the early 1900s, which appears to have been more severe in the eastern part of the study region, coincides with high flows on the North Saskatchewan, South Saskatchewan, and Saskatchewan Rivers (Case and MacDonald 2003) and is also seen in PDSI records (both instrumental and reconstructed) from the western United States (Fye et al. 2003).
4. Links with Pacific forcing Precipitation variability in western Canada and the U.S. Pacific Northwest has been linked with known circulation changes in the Pacific Ocean, particularly the El Niño–Southern Oscillation (ENSO) and the Pacific Decadal Oscillation (PDO; Mantua et al. 1997; Gershunov and Barnett 1998; McCabe and Dettinger 1999; Shabbar and Skinner 2004; Bonsal et al. 1993). The influence of these patterns can be seen in the composite maps of moisture anomalies in the instrumental records during the twentieth century (Fig. 8). These composites show that El Niño and positive PDO years are generally drier, whereas La Niña and negative PDO years are generally wetter across the region. More of the records have negative mean anomalies when ENSO and PDO are in phase but the opposite pattern (La Niña, nega-
2859
tive PDO) is not as strong. Sample depth is limited for situations where the two patterns are out of phase, but generally the patterns are more mixed. Composite maps produced using the reconstructed records for the same groupings are very similar (Fig. 8), again demonstrating the skill of the reconstructions. Unfortunately, it is not possible to extend this analysis back over the past 350 yr, as no definitive PDO reconstructions are presently available.5 Drought conditions across much of the United States (particularly the southwest and extending into the Great Plains) have been linked to La Niña conditions in the Pacific Ocean (Cole and Cook 1998; Cole et al. 2002). As mentioned above, La Niña events typically produce wetter-than-normal conditions in the study region (Fig. 8). Cole et al. (2002) provide a list of persistent droughts that affect the continental United States and the ENSO-sensitive southwest. In the discussion that follows, the timing of these droughts is compared with conditions in the southern Canadian Cordillera to determine whether widespread anomalies of the same sign occur across western North America and if such anomalies coincide with documented La Niña events. As one would expect (based on the ENSO response of the two regions described above), there are instances when conditions in the two regions are opposite (e.g., during the persistent U.S. droughts of the early 1750s, 1780s, 1820s, and the 1950s; see Figs. 4 and 7). There are other periods when persistent drought in the United States coincides with near-normal conditions in the southern Canadian Cordillera (e.g., 1855–65). This is not unexpected as the ENSO teleconnection weakens northward; there is great variability in characteristics of different ENSO events, and ENSO teleconnections to surface climate can be unstable as demonstrated by Cole and Cook (1998) for the continental United States. Sometimes, however, persistent drought affects much of the United States and the southern Canadian Cordillera concurrently (e.g., the 1840s, 1890s, and 1930s; Fig. 6). While it is beyond the scope of this paper to explore the causes of such widespread drought, it is interesting to note that the droughts of the 1840s and 1930s are not particularly strong in the ENSO-sensitive region (i.e., the U.S. Southwest) and the concurrent drought of the 1890s does not correspond with La Niña conditions (Cole et al. 2002). These comparisons indicate that widespread droughts (i.e., droughts that affect
5 Several tree-ring-based reconstructions of the PDO have been developed (e.g., Gedalof and Smith 2001; D’Arrigo et al. 2001; Biondi et al. 2001), but they often disagree with one another prior to the twentieth century.
2860
JOURNAL OF CLIMATE
VOLUME 18
FIG. 8. Composites of instrumental and reconstructed precipitation and PDSI for sites in the southern Canadian Cordillera and northwestern United States over the twentieth century for (a) El Niño and La Niña years, (b) positive and negative PDO years, and (c) in- and out-of-phase ENSO/PDO pairings (Gershunov and Barnett 1998). The instrumental and reconstructed conditions are mapped in standard deviation units from the period 1920–94 (note that some records do not span the entire interval). El Niño and La Niña years are identified using the bivariate ENSO index (Smith and Sardeshmukh 2000) for 1871–2001. Nov–Mar PDO data for 1900–2003 are available online at http://tao.atmos.washington.edu/pdo/.
the western United States and the southern Canadian Cordillera) appear to occur in the absence of La Niña events that are typical of the twentieth century (i.e., something other than La Niña appears to have forced these very large-scale droughts). It is possible that their synchronous occurrence is a coincidence. However, the persistence and scale of these droughts may be related to positive feedbacks that enhance initial land surface anomalies or to more complex interactions with longterm modes of atmospheric/oceanic variability [e.g., the
Atlantic multidecadal oscillation (AMO); Enfield et al. 2001; Gray et al. 2003].6
6
Composite maps produced for the different phases of the AMO show weak but generally consistent anomalies across the southern Cordillera (wet with the positive AMO and dry with the negative AMO), which is consistent with recent work that shows nonsignificant but generally positive correlations between the AMO and summer moisture availability over the region (Shabbar and Skinner 2004).
1 AUGUST 2005
WATSON AND LUCKMAN
5. Summary and conclusions This paper has demonstrated (through comparisons of instrumental and reconstructed precipitation anomaly patterns) that a newly developed set of annual precipitation reconstructions (Watson and Luckman 2004) can be combined with data from adjacent areas to explore spatial patterns of preinstrumental moisture conditions in the southern Canadian Cordillera. This is the first time this type of information has been available for the area. Maps of the most spatially extensive dry events reveal that it is not uncommon for anomalies of the same sign to affect the entire study region. Previous studies and instrumental records have documented the severity and duration of the droughts that affected much of southwestern Canada and the United States during the 1920s and 1930s. An examination of preinstrumental dry events in the region suggest that this drought is of unmatched length and severity over the past 350 yr. Shorter intervals of more severely dry conditions include the early 1720s, 1750s, 1790s, late 1860s through early 1870s, and the 1890s. Many of the driest individual years and short dry periods (i.e., ⬍7 yr) are reconstructed for the eighteenth century. The longest, wettest period identified in the reconstructions occurred in the early twentieth century (⬃1898–1917). Most of the dry and wet periods identified are corroborated by the occurrence of synchronous events in independent records from western North America. The results suggest that the most severe extended dry (1917– 36) and wet (1897–1916) intervals of the past 350 yr occurred during the twentieth century, thus indicating that the instrumental record provides a broad range of conditions that is well-suited for the calibration of treering-climate relationships. It is difficult to compare the spatial patterns of the anomalies without also considering their severity (which is not directly comparable—see previous discussion). However, the dry events of the 1720s, 1750s, 1790s, and early twentieth century appear to have been the most extensive, showing strong anomalies throughout the southern cordillera and the adjacent United States. Extensive negative anomalies also prevailed during the 1890s but appear to have been more intense in the more easterly sites. These results indicate that although there appear to be analogs to the twentiethcentury drought in terms of pattern and severity in the study region, none of the earlier events were as prolonged as that event. The dry periods in the early 1700s and late 1760s–early 1770s are similar in that the dry conditions are concentrated in the interior Canadian sites, whereas the U.S. PDSI grid points actually show wetter conditions. The dry period identified from 1968–
2861
79 is unusual in that many of the more northern and coastal PDSI grid points indicate wetter-than-normal conditions. Overall, the wet periods that have been identified are more regionally coherent than the dry intervals. The extensive wet periods identified for 1778–82, 1879–88, 1898–1916, and 1990–94 are similar in that they show very strong anomalies in eastern portions of the study region and generally weaker anomalies in the United States and interior British Columbia (particularly the first two intervals). In contrast, the strongest anomalies during the 1749–55 and 1980–84 wet intervals are found west of the Rocky Mountains. The three intervals 1689– 1700, 1801–14, and 1819–30 have a number of features in common; they show strong anomalies in both southwestern British Columbia and southwestern Alberta and weak negative anomalies in south-central British Columbia and the southeast corner of the study region. Based on these results, the twentieth-century wet periods do not appear to be unprecedented in this region. A comparison of the occurrence of widespread drought (i.e., drought that is coincident in the study region and the western United States) suggests that these widespread events are not related to La Niña events (a strong control of drought in the southwestern United States). Although this observation is based on only three common drought events, it suggests that caution is warranted when evaluating widespread preinstrumental periods of drought solely in terms of ENSO events [e.g., the sixteenth-century megadrought (Stahle et al. 2000) identified in tree-ring reconstructions from Mexico to western Canada]. The coincident events in the two regions may not be related to a common forcing. However, much more work is needed to evaluate the complex causes and persistence of drought in western North America. In particular, a very limited amount of research has been conducted into relationships between atmospheric and oceanic circulation patterns (and their interactions) and summertime precipitation anomalies (i.e., the period most important to tree growth) in western Canada. Most of the available studies focus on the agriculturally important Canadian Prairies or concentrate on winter conditions. Until a clearer picture is available of the interaction between circulation patterns and moisture anomalies in the southern Canadian Cordillera over the instrumental record, causes of variability in reconstructions of summer precipitation will remain difficult to interpret. Acknowledgments. This research has been funded by the Natural Sciences and Engineering Research Council of Canada, the Meteorological Service of Canada, and most recently a grant from the Canadian Founda-
2862
JOURNAL OF CLIMATE
tion for Climate and Atmospheric Sciences. We thank D. Sauchyn, R. Case, and G. MacDonald for providing their reconstructions, M. Colenutt for contributing data from Waterton Lakes National Park, and Walter Skinner and Xuebin Zhang for their comments on an earlier draft of the manuscript. REFERENCES Biondi, F., A. Gershunov, and D. Cayan, 2001: North Pacific decadal climate variability since AD 1661. J. Climate, 14, 5–10. Bonsal, B. R., A. K. Chakravarti, and R. G. Lawford, 1993: Teleconnections between North Pacific SST anomalies and growing season extended dry spells on the Canadian Prairies. Int. J. Climatol., 13, 865–878. Case, R. A., and G. M. MacDonald, 1995: A dendroclimatic reconstruction of annual precipitation on the western Canadian prairies since A.D. 1505 from Pinus flexilis James. Quat. Res., 44, 267–275. ——, and ——, 2003: Tree ring reconstructions of streamflow for three Canadian prairie rivers. J. Amer. Water Res. Assoc., 39, 703–716. Cole, J. E., and E. R. Cook, 1998: The changing relationship between ENSO variability and moisture balance in the continental United States. Geophys. Res. Lett., 25, 4529–4531. ——, J. T. Overpeck, and E. R. Cook, 2002: Multiyear La Niña events and persistent drought in the contiguous United States. Geophys. Res. Lett., 29, 1647, doi:10.1029/ 2001GL013561. Cook, E. R., D. M. Meko, D. W. Stahle, and M. K. Cleaveland, 1996: Tree-ring reconstructions of past drought across the coterminous United States: Tests of regression method and calibration/verification results. Tree Rings, Environment and Humanity, J. S. Dean, D. M. Meko, and T. W. Swetnam, Eds., Radiocarbon, 155–169. ——, ——, ——, and ——, 1999: Drought reconstructions for the continental United States. J. Climate, 12, 1145–1162. Daniels, L. D., and E. Watson, 2003: Climate–fire–vegetation interactions in the Cariboo forests: A dendrochronological analysis. Rep., Forest Innovation and Investment—Forest Research Program, Vancouver, BC, Canada, 56 pp. D’Arrigo, R., R. Villalba, and G. Wiles, 2001: Tree-ring estimates of Pacific decadal climate variability. Climate Dyn., 18, 219– 224. Enfield, D. B., A. M. Mestas-Nuñez, and P. J. Trimble, 2001: The Atlantic multidecadal oscillation and its relation to rainfall and river flows in the continental U.S. Geophys. Res. Lett., 28, 2077–2080. Filmon, G., cited 2004: Firestorm 2003 Provincial Report. [Available online at http://www.2003firestorm.gov.bc.ca/.] Fye, F. K., D. W. Stahle, and E. R. Cook, 2003: Paleoclimatic analogs to twentieth-century moisture regimes across the United States. Bull. Amer. Meteor. Soc., 84, 901–909. Gedalof, Z., and D. J. Smith, 2001: Interdecadal climate variability and regime-scale shifts in Pacific North America. Geophys. Res. Lett., 28, 1515–1518. Gershunov, A., and T. P. Barnett, 1998: Interdecadal modulation of ENSO teleconnections. Bull. Amer. Meteor. Soc., 79, 2715– 2725. Gray, R., B. Blackwell, K. Iverson, C. Wong, and K. MacKenzie, 2002: Past fire regimes in the interior Douglas fir, dry cool subzone, Fraser variant (IDFdk3). Rep., Lignum, Ltd., 207 pp.
VOLUME 18
Gray, S. T., J. L. Betancourt, C. L. Fastie, and S. T. Jackson, 2003: Patterns and sources of multidecadal oscillations in droughtsensitive tree-ring records from the central and southern Rocky Mountains. Geophys. Res. Lett., 30, 1316, doi:10.1029/ 2002GL016154. Hu, Q., and G. D. Wilson, 2000: Effects of temperature anomalies on the Palmer Drought Severity Index in the central United States. Int. J. Climatol., 20, 1899–1911. Koshida, G., 1992: About drought in Canada. Canadian Climate Program, Environment Canada Publication, Downsview, ON, Canada, 41 pp. Luckman, B. H., 1998: Landscape and climate change in the central Canadian Rockies during the 20th century. Can. Geogr., 42, 319–336. ——, K. R. Briffa, P. D. Jones, and F. H. Schweingruber, 1997: Tree-ring based reconstruction of summer temperatures at the Columbia Icefield, Alberta, Canada, A.D. 1073–1983. Holocene, 7, 375–389. Mantua, N. J., S. R. Hare, U. Zhang, J. M. Wallace, and R. C. Francis, 1997: A Pacific interdecadal climate oscillation with impacts on salmon production. Bull. Amer. Meteor. Soc., 78, 1069–1079. McCabe, G. J., and M. D. Dettinger, 1999: Decadal variations in the strength of ENSO teleconnections with precipitation in the western United States. Int. J. Climatol., 19, 1399–1410. Nemanishen, W., 1998: Drought in the Palliser Triangle: A Provisional Primer. Prairie Farm Rehabilitation Committee, Agriculture and Agri-Food Canada, 58 pp. Oladipo, E. O., 1985: A comparative performance analysis of three meteorological drought indices. J. Climatol., 5, 655–664. Palliser, J., 1895: Papers relative to the exploration of that portion of British North America which lies between the northern branch of the River Saskatchewan and the frontier of the United States; and between the Red River and the Rocky Mountains. Report presented to both Houses of Parliament by command of Her Majesty. Palmer, W. C., 1965: Meteorological drought. Weather Bureau Research Paper 45, U.S. Department of Commerce, 58 pp. Phillips, D., 1990: The Climates of Canada. Minister of Supply and Services Canada, 176 pp. Richman, M. B., 1986: Rotation of principal components. J. Climatol., 6, 293–335. Sauchyn, D. J., and A. B. Beaudoin, 1998: Recent environmental change in the southwestern Canadian Plains. Can. Geogr., 42, 337–353. ——, and W. R. Skinner, 2001: A proxy record of drought severity for the southwestern Canadian Plains. Can. Water Resour. J., 26, 253–272. Shabbar, A., and W. R. Skinner, 2004: Summer drought patterns in Canada and the relationship to global sea surface temperatures. J. Climate, 17, 2866–2880. Smith, C. A., and P. Sardeshmukh, 2000: The effect of ENSO on the intraseasonal variance of surface temperature in winter. Int. J. Climatol., 20, 1543–1557. Stahle, D. W., E. R. Cook, M. K. Cleaveland, M. D. Therrell, D. M. Meko, H. D. Grissino-Mayer, E. Watson, and B. H. Luckman, 2000: Tree-ring data document 16th century megadrought over North America. Eos, Trans. Amer. Geophys. Union, 81, 121, 125. St. George, S., and E. Nielsen, 2003: Palaeoflood records for the Red River, Manitoba, Canada, derived form anatomical treering signatures. Holocene, 13, 547–555.
1 AUGUST 2005
WATSON AND LUCKMAN
Watson, E., and B. H. Luckman, 2001a: Dendroclimatic reconstruction of precipitation for sites in the Canadian Rockies. Holocene, 11, 203–213. ——, and ——, 2001b: The development of a moisture-stressed tree-ring chronology network for the southern Canadian Cordillera. Tree-Ring Res., 57, 149–168. ——, and ——, 2002: The dendroclimatic signal in Douglas-fir and ponderosa pine tree-ring chronologies from the southern Canadian Cordillera. Can. J. For. Res., 32, 1858–1874. ——, and ——, 2004: Tree-ring based reconstructions of precipitation for the southern Canadian Cordillera. Climate Change, 65, 209–241. Wilson, R. J. S., and B. H. Luckman, 2003: Dendroclimatic recon-
2863
struction of maximum summer temperatures from upper treeline sites in interior British Columbia, Canada. Holocene, 13, 851–861. Wolfe, S. A., D. J. Huntley, P. P. David, J. Ollerhead, D. J. Sauchyn, and G. M. MacDonald, 2001: Late 18th century drought-induced sand dune activity, Great Sand Hills, Saskatchewan. Can. J. Earth Sci., 38, 105–117. Woodhouse, C. A., and J. T. Overpeck, 1998: 2000 years of drought variability in the central United States. Bull. Amer. Meteor. Soc., 79, 2693–2712. ——, J. J. Lukas, and P. M. Brown, 2002: Drought in the western Great Plains, 1845–1856: Impacts and implications. Bull. Amer. Meteor. Soc., 83, 1485–1493.
