Statistical Classification of Biosignature Information: Combining Elemental, Molecular, Reflectance, and Raman Data to Increase Life Detection Confidence

Planetary exploration missions seeking past or present signs of life carry not just a single instrument, but a suite. There is a need to study how these multiple data types can be combined to create “composite” biosignatures [1]. Algorithmic methods using existing data on living and non-living systems, though limited by the n = 1 of Earth, can nonetheless be informative. We assembled a database of 1277 measurements spanning 16 representative systems either indicative or non-indicative of life. Five classification (machine learning) methods were used on each individual data type, then on the entire set. This abstract summarizes the results; the data is described in more detail in [2], and methods in [3].