Published online by Cambridge University Press: 20 January 2017
Canonical correlation analysis provides a means of reconstructing quantitative time series of past climatic variables during the last 15,000 years from fossil pollen spectra collected at three sites in the northern Midwest. This multivariate statistical technique was applied to a spatial array of modern pollen and climatic data in order to derive a set of mathematical transfer functions. These transform the fossil pollen spectra directly into quantitative estimates of past climatic values. The basic sequence of climatic events that is reconstructed is in general agreement with previous studies of postglacial climates in the Midwest, but quantitative estimates for certain of the variables, e.g., temperature and precipitation, are given for the first time.
Fossil pollen from three cores collected from lakes in Wisconsin and Minnesota allow a preliminary reconstruction of past east-west and north-south gradients of the climatic variables. Because changes in the circulation patterns in midlatitudes are the principal mechanism causing fluctuations in temperature and precipitation, past records of the atmospheric circulation are reconstructed along with records of temperature and rainfall. The time series derived show that the most pronounced climatic change indicated in Wisconsin and Minnesota occurred at the end of the Pleistocene (the beginning of the Holocene). This change is particularly evident in the climatic variables related to temperature, which rose ca. 3.3°C. A decrease in snowfall also occurred.
During the Holocene, the most marked change appears in the results from Kirchner Marsh, where the amount of dry western air began to increase and the precipitation to decrease about 9500 B.P. A reversal of these changes occurred about 5000 B.P. In contrast, relatively little change occurred among the reconstructed values from Wisconsin; a marked east-west precipitation gradient, therefore, developed in this region of the Midwest from 9500 to 5000 B.P.