Transfer functions under no-analog conditions: Experiments with Indian Ocean planktonic Foraminifera

This paper documents and investigates an important source of inaccuracy when paleoecological equations calibrated on modern biological data are applied downcore: fossil assemblages for which there are no modern analogs. Algebraic experiments with five calibration techniques are used to evaluate the sensitivity of the methods with respect to no-analog conditions. The five techniques are: species regression; principal-components regression [e.g., Imbrie, J., and Kipp, N. G. (1971). In “The Late Cenozoic Ages,” 71–181]; distance-index regression [Hecht, A. D. (1973). Micropaleontology19, 68–77]; diversity-index regression (Williams, D. F., and Johnson, W. C. (1975). Quaternary Research5, 237–250]; weighted-average method [Jones, J. I. (1964). Unpublished Ph. D. Thesis, Univ. of Wisconsin]. The experiments indicate that the four regression techniques extrapolate under no-analog conditions, yielding erroneous estimates. The weighted-average technique, however, does not extrapolate under no-analog conditions and consequently is more accurate than the other techniques. Methods for recognizing no-analog conditions downcore are discussed, and ways to minimize inaccuracy are suggested. Using several equations based on different calibration techniques is recommended. Divergent estimates suggest that no-analog conditions occur and that estimates are unreliable. The value determined by the weighted-average technique, however, may well be the most accurate.