Modern Analog Techniques

Abstract

Modern analog technique, often abbreviated as MAT, is a popular type of transfer function used in paleoceanography to reconstruct past ocean properties from the composition of fossil assemblages extracted from deep-sea sediments. In the vast majority of cases, MAT is applied to reconstruct past sea-surface temperature from remains of marine unicellular microplankton such as planktonic foraminifera, radiolaria, diatoms, and dinoflagellates. The method has been introduced to paleoceanography by Hutson (1980). Mathematically, it is akin to k-nearest neighbor regression. Its procedure is computationally and conceptually very simple. First, a similarity coefficient is calculated between the composition of a fossil assemblage in a geological sample and the compositions of corresponding assemblages in a “calibration dataset” of surface-sediment samples. Then, the surface-sediment samples are sorted by their similarity coefficient, and a subset of those with most similar assemblages, termed the “nearest” or “best modern analogs,” is used in the second step to calculate an estimate of the desired ocean property. The estimate represents a mean of the presentday values of the ocean property at the site of the deposition of the best modern analogs, typically weighted by the respective similarity coefficient of each sample. Like all paleoceanographic transfer functions, MAT relies on the assumption that the reconstructed ocean property is a statistically important determinant of the assemblage composition of the studied organisms. It also assumes that the fossil assemblage and its covariance with the ocean property are represented in the calibration dataset. If this is not the case, the fossil assemblage is said to represent a no-analog condition, which may render the reconstruction of past ocean properties invalid. MAT is a strict interpolator; it does not attempt to extract a general relationship between assemblage composition and environmental factors. Its applicability is therefore contingent on appropriate coverage of the environmental gradient and of the associated assemblage compositions. Because of no-analog conditions due to evolution of niche requirements of the constituent species, assemblage-based transfer functions are typically limited in their applicability to late Quaternary sediments. Sea-surface temperature reconstructions via MATcan be obtained to a precision of 1 C, and the variance among the SST values of the best analogs can be used to assess the uncertainty of a fossil estimate.