Prediction of water in asteroids from spectral data shortward of 3 mu m

Spectra of many asteroids display a 3 mu m absorption feature that has been associated with the presence of water of hydration in clays or hydrated salts. Detection of this feature, however, is difficult through the Earth's atmosphere for various reasons. Correlations were sought and detected between the 3 mu m absorption band and features shortward of 3 mu m, which enabled us to construct a tool for the prediction of water in asteroids from the shorter wavelength part of the spectrum. Such a prediction tool can help concentrate observing resources to those objects most likely to have water. Artificial neural network techniques were used for the investigation of the above correlations and for the prediction of the hydration state of objects, We can predict the presence of water from the <3 mu m spectral window with a high, approximately 90% success rate for spectra for which prediction is possible. However, for about half the spectra in this study, no decision could be made, for lack of sufficient training information. We expect this situation to improve steadily as new data become available. (C) 1997 Academic Press.