An autonomous system has been developed to identify the presence of carbonates in near-IR reflectance spectra (0.35--2.5 μm). This system consists of a set of feature-extraction algorithms that operate in conjunction with a rule-based system to identify carbonates on the basis of presence and characteristics of absorption bands near 2.33 and 2.5 μm. This autonomous system was tested during the February 1999 Marsokhod field operations. While performance was compromised to some extent by noise problems associated with the spectrometer, the autonomous system successfully rejected all noisy spectra. Among the nonnoisy spectra that it accepted, it identified carbonates with a success rate of 70% compared with the performance of the field team, a success rate of 82--90% compared with the performance of the remote science team, and a false positive rate of 0%. ¿ 2001 American Geophysical Union