Lake sediments record large-scale shifts in moistureregimes across the northern prairies of NorthAmerica during the past two millenniaKathleen R. Laird*, Brian F. Cumming*†, Sybille Wunsam‡§, James A. Rusak‡¶, Robert J. Oglesby储, Sherilyn C. Fritz**,and Peter R. Leavitt‡*Paleoecological Environmental Assessment and Research Laboratory, Department of Biology, Queen’s University, Kingston, ON, Canada K7L 3N6;‡Limnology Laboratory, Department of Biology, University of Regina, Regina, SK, Canada S4S 0A2;储National Space Science and TechnologyCenter, Marshall Space Flight Center, National Aeronautics and Space Administration, 320 Sparkman Drive, Huntsville, AL 35805; and**Department of Geosciences, University of Nebraska, Lincoln, NE 68588Communicated by H. E. Wright, Jr., University of Minnesota, Minneapolis, MN, January 10, 2003 (received for review May 14, 2002)Six high-resolution climatic reconstructions, based on diatom anal-yses from lake sediment cores from the northern prairies of NorthAmerica, show that shifts in drought conditions on decadalthrough multicentennial scales have prevailed in this region for atleast the last two millennia. The predominant broad-scale patternseen at all sites is a major shift in moisture regimes from wet to dry,or vice versa (depending on location), that occurred after a periodof relative stability. These large-scale shifts at the different sitesexhibit spatial coherence at regional scales. The three Canadiansites record this abrupt shift between anno Domini 500 and 800,and subsequently conditions become increasingly variable. Allthree U.S. sites underwent a pronounced change, but the timing ofthis change is between anno Domini 1000 and 1300, thus later thanin all of the Canadian sites. The mechanisms behind these patternsare poorly understood, but they are likely related to changes in theshape and location of the jet stream and associated storm tracks.If the patterns seen at these sites are representative of the region,this observed pattern can have huge implications for future wateravailability in this region.Drought is a recurring natural feature of climate that has haddramatic environmental, economic, and social impacts onmodern (1) and ancient (2) civilizations. Decade- to century-scale episodes of prolonged drought or high rainfall have beenrecorded from North America in continental archives as diverseas tree rings (3), tree stumps (4), lake sediments (5), and riverdeposits (6, 7). The prairie region of North America is partic-ularly susceptible to extreme droughts (8). However, few high-resolution proxy records of climate exist from the prairies.Tree-ring records on the prairies typically are spatially limited towooded areas at the periphery of the prairies (9) and temporallylimited to, at best, ⬇500 years (10). Sediments from closed-basinlakes can provide high-temporal-resolution paleoclimatic infor-mation from prairie regions for much longer periods.Here we provide evidence from high-resolution (subdecadalto decadal) sediment core records from six lakes on the Cana-dian and northern U.S. prairies (Fig. 1). All of the sites indicatethat shifts in drought regimes have been a prevalent feature ofthis region, occurring on decadal through multicentennial scales.Inferred changes in climatic conditions over the past two mil-lennia are based on analysis of diatom assemblages preserved insediment cores, a commonly used technique for tracking pastclimatic conditions (11). Our focus here is on long-term dynam-ics and broad-scale similarities among the lake records. Thisapproach is in part undertaken because of the inherent difficul-ties of comparing short-term dynamics across records con-strained by carbon-dated chronologies.Materials and MethodsDiatom Inferences. Diatom remains in sediment cores from sixlakes were used to reconstruct variables influenced by climaticconditions. Diatom-inferred salinity estimates were used in fivelakes to reconstruct relative changes in effective moisture [E兾P,the balance between precipitation (P) and evaporation (E)], withhigh salinity implying high E兾P (12). Diatoms are common algaeof inland saline and freshwater lakes, and their taxonomicdistribution is highly related to lakewater salinity (13, 14).Statistical models developed from the contemporary distributionof diatom taxa in various regions of the world (reviewed in ref.12) have been used successfully to estimate past lakewatersalinity from diatom assemblages preserved in well-dated sedi-ment cores (5, 15, 16). The predictive ability of diatom-basedsalinity-inference models, as judged from the relationship be-tween inferred and observed salinity in a suite of modern lakes,is strong and highly significant (bootstrapped r2⫽ 0.8–0.9, P ⬍0.01; ref. 12). To infer past lakewater salinity in this study, wedeveloped a modern-day calibration set of 287 lakes, consistingof 79 lakes from the prairie region (ref. 14; S.C.F., unpublisheddata) and 208 lakes from British Columbia (13). The resultantpredictive model to infer salinity was as strong as earlier models,but the larger number of lakes (and taxa) on which this modelwas based allowed an expanded set of appropriate analogs forreconstructing salinity from the diatom assemblages encoun-tered in the cores.Total phosphorus (TP) was inferred from diatoms in a sedi-ment core from Elk Lake, because changes in phosphorus werethe dominant signal in the diatom assemblages. In Elk Lake,TP levels are driven largely by the degree of water-columnmixing and stratification as related to general climatic conditions(17). Thus TP estimates at Elk Lake can provide a sensitive proxyfor climate. Diatoms are highly influenced by TP levels, whichcan be reconstructed by using inference models (reviewed inref. 18).Correspondence analysis, which indicates the main directionof variation in multivariate data, such as diatom assemblages ina core, was undertaken for each of the six sites to ensure that ourinferences tracked the major changes in the diatom assemblages.Correspondence analysis axis-one scores for four of the lakeswere strongly correlated with inferred log-salinity estimates (r ⫽0.67–0.94, P ⬍ 0.01), and for Elk Lake, with log TP (r ⫽ 0.90,P ⬍ 0.01). For Nora Lake, inferred salinity was strongly relatedto axis-two scores (r ⫽ 0.87, P ⬍ 0.01). These strong correlationsconfirm that our diatom-based salinity or TP inferences are agood summary and simplification of the changes in the