In control theory, an observer is a numerical algorithm which estimates unmeasured physical variables from those that are measured with the help of dynamical models that correlate them. This paper presents nonlinear observer designs for the catalytic partial oxidation (CPO), water gas shift (WGS) and preferential oxidation (PROX) reactors in fuel cell power systems. The observers make use of temperature measurements (and possibly one more variable, such as pressure) to estimate the mole fractions of each species in the reactors. Estimation of hydrogen and carbon monoxide fractions is particularly useful for monitoring and control of the fuel cell system. An advantage of our designs is that they are based on the reaction invariants and do not rely on a knowledge of reaction rate expressions. (C) 2005 International Association for Hydrogen Energy. Published by Elsevier Ltd. All rights reserved.