Application of artificial neural networks to paleoceanographic data

Abstract

Abstract Artificial neural networks are computer systems that have the ability to ‘learn’ a set of output, or target, vectors from a set of input vectors. Learning is achieved by self-adjustment of a set of parameters to minimize the error between a desired output and the actual network output. We have explored the potential of this approach in paleoceanography by application of a neural network algorithm to a problem involving prediction of sea surface-water temperatures from relative abundances of planktonic foraminifer species in the southern Indian Ocean. We employed a backpropagation (BP) network to assess how well it was able to predict the actual summer and winter surface-water temperatures. We compared the results with those obtained from statistical methods previously used for temperature predictions: Imbrie-Kipp Transfer Functions, the Modern Analog Technique, and Soft Independent Modelling of Class Analogy. The efficiency of predictions was tested using the Leaving One Out technique in which each of the observations in the data set is left out one at a time, while the remaining observations are used to generate a predictor. The accuracy of the predictor is then tested on the observation left out by comparison with its actual value. A set of tests using 1, 2, 3, 4, 5, and 10 neurons (processing elements) in a 3-layer BP network showed that a network with 3 neurons gave the smallest errors of prediction for both summer and winter temperatures, 0.71 and 0.76, respectively. Corresponding errors for the statistical pattern-recognition techniques ranged between 1.01 and 1.26 for summer temperatures and 1.05-1.13 for winter temperatures. Hence, predictions of paleotemperatures from new data on planktonic foraminifer relative abundances in the southern Indian Ocean may be made with a precision of ±0.7-0.8°C using the BP network and ±1.0–1.3°C using the statistical pattern-recognition procedures. The BP network was thus the most successful among the methods employed here for temperature predictions. Artificial neural networks may, therefore, be seen as a viable alternative to more conventional approaches to data analysis in paleoceanography.